0% found this document useful (0 votes)

89 views124 pages

HH PDF

The HH Package provides functions for statistical analysis and data display. It supports the book Statistical Analysis and Data Display by Heiberger and Holland from Springer. The package contains functions for ANOVA, ANCOVA, residuals analysis, graphical displays, and more. It is maintained by Richard Heiberger and depends on packages like lattice, stats, and grid for graphics and statistical analysis capabilities.

Uploaded by

learning English

We take content rights seriously. If you suspect this is your content, claim it here.

Available Formats

Download as PDF, TXT or read online on Scribd

0% found this document useful (0 votes)

89 views124 pages

HH PDF

Uploaded by

learning English

We take content rights seriously. If you suspect this is your content, claim it here.

Available Formats

Download as PDF, TXT or read online on Scribd

You are on page 1/ 124

The HH Package

August 7, 2007
Type Package

Title Statistical Analysis and Data Display: Heiberger and Holland

Version 2.1-3

Date 2007-07-29

Author Richard M. Heiberger

Maintainer Richard M. Heiberger <rmh@temple.edu>

Depends R (>= 2.5.0), lattice, stats, grid, multcomp (>= 0.991-7), mvtnorm, methods

Description Support software for Statistical Analysis and Data Display (Springer, ISBN
0-387-40270-5). This contemporary presentation of statistical methods features extensive use of
graphical displays for exploring data and for displaying the analysis. The authors demonstrate
how to analyze data—showing code, graphics, and accompanying computer listings—for all the
methods they cover. They emphasize how to construct and interpret graphs, discuss principles of
graphical design, and show how accompanying traditional tabular results are used to confirm the
visual impressions derived directly from the graphs. Many of the graphical formats are novel and
appear here for the first time in print. All chapters have exercises.

License GPL version 2 or newer

R topics documented:
GSremove . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
HH-package . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
X.residuals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
ae.dotplot . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
ancova-class . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
ancova . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
anova.mean . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
aov.sufficient . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
arima.diag.hh . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
arma.loop . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
as.multicomp . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19

1
2 R topics documented:

axis.i2wt . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
bwplot.formula . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
ci.plot . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
col.hh . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
diag.maybe.null . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
do.formula.trellis.xysplom . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
export.eps . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
extra . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
gof.calculation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
grid.yaxis.hh . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
hh . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
hov . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
if.R . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
interaction.positioned . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
interaction2wt . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
intxplot . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
ladder . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
legendGrob2wt . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
lm.case . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
mcalinfct . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
mmc . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
mmc.mean . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
multicomp.order . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
multicomp.reverse . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
norm.curve . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
npar.arma . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
objip . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
odds.ratio . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
orthog.complete . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
panel.acf . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
panel.bwplot.intermediate.hh . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
panel.bwplott . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
panel.cartesian . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
panel.ci.plot . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
panel.dotplot.tb . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
panel.interaction2wt . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
panel.pairs.hh . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
panel.xysplom . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
partial.corr . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
plot.hov . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
plot.mmc.multicomp . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
plot.multicomp . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
position . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
positioned-class . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
print.tsdiagplot . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
push.vp.hh . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
regr1.plot . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
resid.squares . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
GSremove 3

residual.plots . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
seqplot . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
seqplot.forecast . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
strip.background0 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
strip.xysplom . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
sufficient . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
summary.arma.loop . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
t.trellis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
tsacfplots . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
tsdiagplot . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
vif . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
xysplom . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116

Index 121

GSremove Remove selected GraphSheetPages in the S-Plus Windows GUI

Graphsheet

Description
Remove selected GraphSheetPages in the S-Plus Windows GUI Graphsheet. This does the same
task as right-click/delete on the tabs of the GraphSheet.

Usage
GSremove(pages, sheet = "GSD2$Page")

Arguments
pages Page numbers in the tabs at the bottom of the Graphsheet.
sheet Defaults to GSD2, the first name that is used when the graphsheet or trellis.device
function is used.

Author(s)
Richard M. Heiberger <rmh@temple.edu>

X.residuals Residuals from the regression of each column of a data.frame against

all the other columns.

Description
Calculate the residuals from the regression of each column of a data.frame against all the other
columns.

Usage
X.residuals(x, ...)

## Default S3 method:
X.residuals(x, y.name, na.action = na.exclude, ...) ## x is a data.frame

## S3 method for class 'formula':

X.residuals(x, data, na.action = na.exclude, ...) ## x is a formula

## S3 method for class 'lm':

X.residuals(x, na.action = na.exclude, ...) ## x is a "lm" object computed with x=

Arguments
x data.frame, or formula, or lm object computed with x=TRUE.
na.action See na.action.
... additional arguments.
y.name Name of Y-variable to be excluded from the computations.
data A data frame in which the variables specified in the formula will be found. If
missing, the variables are searched for in the standard way.

Value
Data.frame of residuals, one column from each regression.

Author(s)
Richard M. Heiberger <rmh@temple.edu>
6 ae.dotplot

See Also
lm in R or lm in S-Plus, vif, lm.case.

Examples
usair <- read.table(hh("datasets/usair.dat"),
col.names=c("SO2","temp","mfgfirms","popn",
"wind","precip","raindays"))
usair$lnSO2 <- log(usair$SO2)
usair$lnmfg <- log(usair$mfgfirms)
usair$lnpopn <- log(usair$popn)

usair.lm <- lm(lnSO2 ~ temp + lnmfg + wind + precip, data=usair, x=TRUE)

X.residuals(usair.lm) ## the lm object must be computed with x=TRUE

X.residuals(lnSO2 ~ temp + lnmfg + wind + precip, data=usair)

X.residuals(usair)

X.residuals(usair, y.name="lnSO2")

ae.dotplot AE (Adverse Events) dotplot of incidence and relative risk

Description
A two-panel display of the most frequently occurring AEs in the active arm of a clinical study. The
first panel displays their incidence by treatment group, with different symbols for each group. The
second panel displays the relative risk of an event on the active arm relative to the placebo arm, with
95% confidence intervals for a 2×2 table. The AEs are ordered by relative risk so that events with
the largest increases in risk for the active treatment are prominent at the top of the display.

Usage
ae.dotplot(xr,
A.name = levels(xr$RAND)[1], B.name = levels(xr$RAND)[2],
col.AB = c("red","blue"), pch.AB = c(16, 17),
main.title = "Most Frequent On-Therapy Adverse Events Sorted by Relative
main.cex = 1,
cex.AB.points = NULL, cex.AB.y.scale = 0.6,
position.left = c(0, 0, 0.7, 1), position.right = c(0.61, 0, 0.98, 1),
ae.dotplot 7

key.y = -0.2, CI.percent=95)

logrelrisk(ae, A.name, B.name, crit.value=1.96)

panel.ae.leftplot(x, y, groups, col.AB, ...)

panel.ae.rightplot(x, y, ..., lwd=6, lower, upper)

panel.ae.dotplot(x, y, groups, ..., col.AB, pch.AB, lower, upper) ## R only

Arguments
ae data.frame containing the first 4 columns of xr described below.
xr RAND: treatment as randomized (factor).\ PREF: adverse event symptom name
(factor).\ SN: number of patients in treatment group.\ SAE: number of patients
in each group for whom the event PREF was observed.\ PCT: SAE/SN as a
percent.\ relrisk: Relative risk defined as PCT for the B treatment divided
by PCT for the A treatment.\ logrelrisk: natural logarithm of relrisk.\
ase.logrelrisk: asymptotic standard error of logrelrisk.\ logrelriskCI.lower,
logrelriskCI.upper: confidence interval for logrelrisk. \ relriskCI.lower,
relriskCI.upper: back transform of the CI for the log relative risk into the
relative risk scale.\
A.name, B.name
Names of treatment groups (in x$RAND).
col.AB, pch.AB, cex.AB.points
color, plotting character and character expansion for the individual points on the
left plot.
cex.AB.y.scale
Character expansion for the left tick labels (the symptom names).
main.title, main.cex
Main title and character expansion for the combined plot in ae.dotplot.
position.left, position.right
position of the left and right plots. This argument is use in S-Plus only, not in
R. See the discussion of position in print.trellis in R, print.trellis
in S-Plus.
key.y Position of the key (legend) in the combined plot. This is the y argument of the
key. See key in S-Plus, and the discussion of the key argument to xyplot
in xyplot in R.
crit.value Critical value used to compute confidence intervals on the log relative risk.
Defaults to 1.96. User is responsible for specifying both crit.value and
CI.percent consistently.
CI.percent Confidence percent associated with the crit.value Defaults to 95. User is
responsible for specifying both crit.value and CI.percent consistently.
x, y, groups, lwd, ...
standard panel function arguments.
lower, upper
xr$logrelriskCI.lower and xr$logrelriskCI.upper inside the panel functions.
8 ae.dotplot

Details
The second panel shows relative risk of an event on the active arm (treatment B) relative to the
placebo arm (treatment A), with 95% confidence intervals for a 2×2 table. Confidence intervals on
the log relative risk are calculated using the asymptotic standard error formula given as Equation
3.18 in Agresti A., Categorical Data Analysis. Wiley: New York, 1990.

Value
logrelrisk takes an input data.frame of the form x described in the argument list and returns a
data.frame consisting of the input argument with additional columns as described in the argument
xr. The result column of symptom names PREF is an ordered factor, with the order specified by
the relative risk.
ae.leftplot returns a "trellis" object containing a horizontal dotplot of the percents against
each of the symptom names.
ae.rightplot returns a "trellis" object containing a horizontal plot on the log scale of the
relative risk confidence intervals against each of the symptom names.
ae.dotplot calls both ae.leftplot and ae.rightplot and combines their plots into a
single display with a single set of left axis labels, a main title, and a key. The value returned
invisibly is a list of the full left trellis object and the right trellis obect with its left labels blanked
out. Printing the value will not usually be interesting as the main title and key are not included. It is
better to call ae.dotplot directly, perhaps with a change in some of the positioning arguments.

Author(s)
Richard M. Heiberger <rmh@temple.edu>

References
Ohad Amit, Lane, Peter W., & Richard M. Heiberger (2007). Graphical Approaches to the Analysis
of Safety Data from Clinical Trials. Pharmaceutical Statistics. http://www3.interscience.
wiley.com/cgi-bin/abstract/114129388/ABSTRACT?CRETRY=1\&SRETRY=0

Examples
## variable names in the input data.frame aeanonym
## RAND treatment as randomized
## PREF adverse event symptom name
## SN number of patients in treatment group
## SAE number of patients in each group for whom the event PREF was observed
##
## Input sort order is PREF/RAND

aeanonym <- read.table(hh("datasets/aedotplot.dat"), header=TRUE, sep=",")

aeanonym[1:4,]

## Calculate log relative risk and confidence intervals (95

## logrelrisk sets the sort order for PREF to match the relative risk.
aeanonymr <- logrelrisk(aeanonym,
A.name=levels(aeanonym$RAND)[1],
ancova-class 9

B.name=levels(aeanonym$RAND)[2])
aeanonymr[1:4,]

## construct and print plot on current graphics device

ae.dotplot(aeanonymr,
A.name="TREATMENT A (N=216)",
B.name="TREATMENT B (N=431)")
## export.eps(h2("stdt/figure/aerelrisk.eps"))
## This looks great on screen and exports badly to eps.
## We recommend drawing this plot directly to the postscript device:
##
## trellis.device(postscript, color=TRUE, horizontal=TRUE,
## colors=ps.colors.rgb[
## c("black", "blue", "red", "green",
## "yellow", "cyan","magenta","brown"),],
## onefile=FALSE, print.it=FALSE,
## file=h2("stdt/figure/aerelrisk.ps"))
## ae.dotplot(aeanonymr,
## A.name="TREATMENT A (N=216)",
## B.name="TREATMENT B (N=431)")
## dev.off()

ancova-class Class "ancova" Analysis of Covariance

Description
Analysis of Covariance. The class is an extension of "aov" and "lm". It is identical to the "aov" for
a single factor and a single covariate plus an attribute which contains a "trellis" object. Four
different models are included in the class. See ancova for the examples.

Objects from the Class

A virtual Class: No objects may be created from it.

Extends
Class "aov", directly. Class "lm", by class "aov", distance 2. Class "mlm", by class "aov",
distance 2, with explicit test and coerce. Class "oldClass", by class "aov", distance 3. Class
"oldClass", by class "aov", distance 4, with explicit test and coerce.

Methods
No methods defined with class "ancova" in the signature. S3-type methods are "anova.ancova",
"coef.ancova", "coefficients.ancova", "model.frame.ancova", "plot.ancova",
"predict.ancova", "print.ancova", "summary.ancova". "plot.ancova(x)"
plots a standard lm plot of x. "print.ancova(x)" prints the anova(x) and the trellis at-
tribute. The remaining methods use NextMethod.
10 ancova

Author(s)
Richard M. Heiberger <rmh@temple.edu>

ancova Compute and plot oneway analysis of covariance

Description

Usage
ancova(formula, data.in = sys.parent(), ...,
x, groups, transpose = FALSE,
display.plot.command = FALSE,
superpose.level.name = "superpose",
ignore.groups = FALSE, ignore.groups.name = "ignore.groups",
blocks, blocks.pch = letters[seq(levels(blocks))],
layout, between, main)

panel.ancova(x, y, subscripts, groups,

transpose = FALSE, ...,
coef, contrasts, classes,
ignore.groups, blocks, blocks.pch, blocks.cex)

## The following are ancova methods for generic functions.

## S3 method for class 'ancova':
anova(object, ...)

## S3 method for class 'ancova':

predict(object, ...)

## S3 method for class 'ancova':

print(x, ...) ## prints the anova(x) and the trellis attribute

## S3 method for class 'ancova':

model.frame(formula, ...)
ancova 11

## S3 method for class 'ancova':

summary(object, ...)

## S3 method for class 'ancova':

plot(x, y, ...) ## standard lm plot. y is always ignored.

## S3 method for class 'ancova':

coef(object, ...)

## S3 method for class 'ancova':

coefficients(object, ...)

Arguments
formula A formula specifying the model.
data.in A data frame in which the variables specified in the formula will be found. If
missing, the variables are searched for in the standard way.
... Arguments to be passed to aov, such as subset or na.action.
x Covariate in ancova, needed for plotting when the formula does not include
x. "aov" object in print.ancova, to match the argument of the print
generic function. Variable to plotted in "panel.ancova".
groups Factor. Needed for plotting when the formula does not include groups after
the conditioning bar "|".
transpose S-Plus: The axes in each panel of the plot are transposed. The analysis is iden-
tical, just the axes displaying it have been interchanged. R: no effect.
display.plot.command
The default setting is usually what the user wants. The alternate value TRUE
prints on the console the command that draws the graph. This is strictly for
debugging the ancova command.
superpose.level.name
Name used in strip label for superposed panel.
ignore.groups
When TRUE, an additional panel showing all groups together with a common
regression line is displayed.
ignore.groups.name
Name used in strip label for ignore.groups panel.
blocks Additional factor used to label points in the panels.
blocks.pch Alternate set of labels used when a blocks factor is specified.
blocks.cex Alternate set of cex used when a blocks factor is specified.
layout The layout of multiple panels. The default is a single row. See details.
between Space between the panels for the individual group levels and the superpose panel
including all groups.
main Character with a main header title to be done on the top of each page.
12 ancova

y,subscripts In "panel.ancova", see panel.xyplot in R and both xyplot and trel-

lis.args in S-Plus.
object An "aov"
coef, contrasts, classes
Internal variables used to communicate between ancova and panel.ancova.
They keep track of the constant or different slopes and intercepts in each panel
of the plot.

Details
The ancova function does two things. It passes its arguments directly to the aov function and
returns the entire aov object. It also rearranges the data and formula in its argument and passes
that to the xyplot function. The trellis attribute is a trellis object consisting of a series
of plots of y ~ x. The left set of panels is conditioned on the levels of the factor groups. The
right panel is a superpose of all the groups.

Value
The result object is an ancova object which consists of an ordinary aov object with an additional
trellis attribute. The default print method is to print both the anova of the object and the
trellis attribute.

Author(s)
Richard M. Heiberger <rmh@temple.edu>

## y ~ x ## constant line across all groups

ancova(Sodium ~ Calories, data=hotdog, groups=Type)

## y ~ a ## different horizontal line in each group

ancova(Sodium ~ Type, data=hotdog, x=Calories)

## This is the usual usage

## y ~ x + a or y ~ a + x ## constant slope, different intercepts
ancova(Sodium ~ Calories + Type, data=hotdog)
ancova(Sodium ~ Type + Calories, data=hotdog)
anova.mean 13

## y ~ x * a or y ~ a * x ## different slopes, and different intercepts

ancova(Sodium ~ Calories * Type, data=hotdog)
ancova(Sodium ~ Type * Calories, data=hotdog)

## y ~ a * x ## save the object and print the trellis graph

hotdog.ancova <- ancova(Sodium ~ Type * Calories, data=hotdog)
attr(hotdog.ancova, "trellis")

## label points in the panels by the value of the block factor

apple <- read.table(hh("datasets/apple.dat"), header=TRUE)
apple$treat <- factor(apple$treat)
contrasts(apple$treat) <- contr.treatment(6)
apple$block <- factor(apple$block)
ancova(yield ~ treat + pre, data=apple, blocks=block)

anova.mean ANOVA table from the group sample sizes, means, and standard devi-
ations.

Description
Oneway ANOVA table from the summary information consisting of group sample sizes, means,
and standard deviations. The full dataset is not needed.

Usage
anova.mean(object, n, ybar, s, ..., ylabel = "ylabel")

Arguments
object level names
n sample size for each level
ybar sample mean for each level
s sample standard deviation for each level
... other arguments (not used)
ylabel name of response variable

Value
Analysis of variance table, identical to the ANOVA table that would have been produced by anova.lm
in S-Plus and anova.lm in R if the original data, rather than the summary data, had been available.

Author(s)
Richard M. Heiberger <rmh@temple.edu>
14 aov.sufficient

See Also
anova.lm in S-Plus and anova.lm in R. plot.mmc.multicomp

Examples
## pulmonary data used in Hsu and Peruggia paper defining the mean-mean plot
## See ?plot.mmc.multicomp for details on the dataset.

pulmonary <- read.table(hh("datasets/pulmonary.dat"), header=TRUE,

row.names=NULL)
names(pulmonary)[3] <- "FVC"
names(pulmonary)[1] <- "smoker"
pulmonary$smoker <- factor(pulmonary$smoker, levels=pulmonary$smoker)
row.names(pulmonary) <- pulmonary$smoker
pulmonary

anova.mean(pulmonary$smoker,
pulmonary$n,
pulmonary$FVC,
pulmonary$s,
ylabel="pulmonary")

aov.sufficient Analysis of variance from sufficient statistics for groups.

Description
Analysis of variance from sufficient statistics for groups. For each group, we need the factor level,
the response mean, the within-group standard deviation, and the sample size. The correct ANOVA
table is produced. The residuals are fake. The generic vcov and summary.lm don’t work for the
variance of the regression coefficients in this case. Use vcov.sufficient.

Usage
aov.sufficient(formula, data = NULL,
projections = FALSE, qr = TRUE, contrasts = NULL,
weights = data$n, sd = data$s,
...)

vcov.sufficient(object, ...)

Arguments
formula, data, projections, qr, contrasts, ...
See aov in R, aov in S-Plus.
weights See lm in R, lm in S-Plus.
sd vector of within-group standard deviations.
object "aov" object constructed by aov.sufficient. It also works with regular
aov objects.
aov.sufficient 15

Value
For aov.sufficient, an object of class c("aov", "lm"). For vcov.sufficient, a function
that returns the covariance matrix of the regression coefficients.

Note
The residuals are fake. They are all identical and equal to the MLE standard error (sqrt(SumSq.res/df.tot)).
They give the right ANOVA table. They may cause confusion or warnings in other programs. The
standard errors and t-tests of the coefficients are not calculated by summary.lm. Using the aov
object from aov.sufficient in glht requires the vcov. and df arguments.

Author(s)
Richard M. Heiberger <rmh@temple.edu>

See Also
MMC and aov in R, aov in S-Plus.

Examples
## This example is from Hsu and Peruggia

## This is the R version

## See ?aov.sufficient for S-Plus

if.R(s={},
r={

pulmonary <- read.table(hh("datasets/pulmonary.dat"), header=TRUE,

row.names=NULL)
names(pulmonary)[3] <- "FVC"
names(pulmonary)[1] <- "smoker"
pulmonary$smoker <- factor(pulmonary$smoker, levels=pulmonary$smoker)
row.names(pulmonary) <- pulmonary$smoker
pulmonary
pulmonary.aov <- aov.sufficient(FVC ~ smoker,
data=pulmonary)
summary(pulmonary.aov)

pulmonary.mca <- glht(pulmonary.aov,

linfct=mcp(smoker="Tukey"),
df=pulmonary.aov$df.residual,
vcov.=vcov.sufficient)
plot(pulmonary.mca)

pulm.lmat <- cbind("npnl-mh"=c( 1, 1, 1, 1,-2,-2), ## not.much vs lots

"n-pnl" =c( 3,-1,-1,-1, 0, 0), ## none vs light
"p-nl" =c( 0, 2,-1,-1, 0, 0), ## {} arbitrary 2 df
"n-l" =c( 0, 0, 1,-1, 0, 0), ## {} for 3 types of light
"m-h" =c( 0, 0, 0, 0, 1,-1)) ## moderate vs heavy
16 arima.diag.hh

dimnames(pulm.lmat)[[1]] <- row.names(pulmonary)

if.R(r=pulm.lmat <- rbind(Int=0, pulm.lmat[-1,]),
s={})
pulm.lmat

pulmonary.mmc <- glht.mmc(pulmonary.aov,

linfct=mcp(smoker="Tukey"),
df=pulmonary.aov$df.residual,
vcov.=vcov.sufficient,
lmat=pulm.lmat,
calpha=attr(confint(pulmonary.mca)$confint,"calpha"))
plot(pulmonary.mmc, print.mca=TRUE, print.lmat=FALSE)

## tiebreaker plot, with contrasts ordered to match MMC plot,

## with all contrasts forced positive and with names also reversed,
## and with matched x-scale.
plot(confint(as.glht(pulmonary.mmc$mca)),
xlim=par()$usr[1:2], xaxs="i",
main="", xlab="")

## orthogonal contrasts
plot(pulmonary.mmc, print.lmat=TRUE, col.lmat.signif='blue', col.iso='gray')

## pairwise and orthogonal contrasts on the same plot

plot(pulmonary.mmc, print.mca=TRUE, print.lmat=TRUE,
col.mca.signif='red', col.lmat.signif='blue', col.iso='gray',
lty.lmat.not.signif=2)

})

arima.diag.hh Repair design error in S-Plus arima.diag

Description
Repair design error in S-Plus arima.diag.

Usage
arima.diag.hh(z, acf.resid = TRUE,
lag.max = round(max(gof.lag + n.parms + 1, 10 * log10(n))),
gof.lag = 15, resid = FALSE,
std.resid = TRUE, plot = TRUE, type = "h", ...,
x=eval(parse(text = series.name), local = sys.parent()))

Arguments
z,acf.resid,lag.max,gof.lag,resid,std.resid,plot,type,...
See arima.diag.
x The time series. This must be specified when arima.diag is called from inside
another function.
arma.loop 17

Details
Repairs design flaw in S-Plus arima.diag. The location of the time series is hardwired one level
up, so it can’t be found when arima.diag is not one level down from the top.
This function is a no-op in R.

Value
See arima.diag.

Author(s)
Richard M. Heiberger <rmh@temple.edu>

series Character string describing the time series.

... Additional arguments for arima.mle or arima.
z For diag.arma.loop, an "arma.loop" object. For rearrange.diag.arma.loop,
an "diag.arma.loop" object.
lag.max Maximum lag for the acf and pacf plots.
gof.lag Maximum lag for the gof plots.

Details

S-Plus and R have different functions, with different input argument names and different compo-
nents in their value.

Value

arma.loop: "arma.loop" object which is a matrix of lists, each containing an arima model.
diag.arma.loop: "diag.arma.loop" object which is a matrix of lists, each containing the
standard diagnostics for one arima model.
rearrange.diag.arma.loop: List of matrices, each containing all the values for a specific
diagnostic measure collected from the set of arima models.

Author(s)

Richard M. Heiberger (rmh@temple.edu)

References

"Displays for Direct Comparison of ARIMA Models" The American Statistician, May 2002, Vol.
56, No. 2, pp. 131-138. Richard M. Heiberger, Temple University, and Paulo Teles, Faculdade de
Economia do Porto.
Richard M. Heiberger and Burt Holland (2004), Statistical Analysis and Data Display, Springer,
ISBN 0-387-40270-5

See Also

tsdiagplot

Examples

## see tsdiagplot for the example

as.multicomp 19

as.multicomp Support functions in R for MMC (mean–mean multiple comparisons)

plots.

Description
MMC plots: In R, functions used to interface the glht in R to the MMC functions designed with
S-Plus multicomp notation. These are all internal functions that the user doesn’t see.

Usage
## S3 method for class 'mmc.multicomp':
print(x, ...)

## S3 method for class 'multicomp':

print(x, ...)

## print.multicomp.hh(x, digits = 4, ..., height=T) ## S-Plus only

## S3 method for class 'multicomp.hh':

print(x, ...) ## R only

print.glht.mmc.multicomp(x, ...) ## R. yes, spell it out.

as.multicomp(x, ...)

## S3 method for class 'glht':

as.multicomp(x, ## glht object
focus, ## currently required
ylabel=as.character(terms(x$model)[[2]]),
means=model.tables(x$model, type="means",
cterm=focus)$tables[[focus]],
lmat=t(x$linfct),
lmat.rows=-1,
lmat.scale.abs2=TRUE,
estimate.sign=1,
order.contrasts=TRUE,
contrasts.none=FALSE,
level=0.95,
calpha=NULL,
method=x$type,
...
)

as.glht(x, ...)

## S3 method for class 'multicomp':

20 as.multicomp

as.glht(x, ...)

Arguments
x "glht" object for as.multicomp. A "mmc.multicomp" object for print.mmc.multicomp
and print.glht.mmc.multicomp. A "multicomp" object for as.glht
and print.multicomp.
... other arguments.
focus name of focus factor.
ylabel response variable name on the graph.
means means of the response variable on the focus factor.
lmat, lmat.rows
mmc
lmat.scale.abs2
logical, almost always TRUE. If it is not TRUE, then the contrasts will not be
properly placed on the MMC plot.
estimate.sign
numeric. 1: force all contrasts to be positive by reversing negative contrasts.
−1: force all contrasts to be negative by reversing positive contrasts. Leave
contrasts as they are constructed by glht.
order.contrasts
logical. If TRUE, order contrasts by height (see MMC).
contrasts.none
logical. This is an internal detail. The “contrasts” for the group means are not
real contrasts in the sense they don’t compare anything. glht.mmc.glht sets
this argument to TRUE for the none component.
level Confidence level. Defaults to 0.95.
calpha User-specified critical point. See confint.glht.hh and confint.glht.
method See type in confint.glht.

Details
The mmc.multicomp print method displays the confidence intervals and heights on the MMC
plot for each component of the mmc.multicomp object.
print.multicomp displays the confidence intervals and heights for a single component.
print.glht.mmc.multicomp(x, ...) uses print.glht on each component of a mmc.multicomp
object and therefore prints only the estimates of the comparisons.

Value
as.multicomp is a generic function to change its argument to a "multicomp" object.
as.multicomp.glht changes an "glht" object to a "multicomp" object. If the model
component of the argument "x" is an "aov" object then the standard error is taken from the
anova(x$model) table, otherwise from the summary(x). With a large number of levels for the
focus factor, the summary(x) function is exceedingly slow (80 minutes for 30 levels on 1.5GHz
Windows XP). For the same example, the anova(x$model) takes a fraction of a second.
axis.i2wt 21

Note
The multiple comparisons calculations in R and S-Plus use completely different libraries. MMC
plots in R are based on glht. MMC plots in S-Plus are based on multicomp. The MMC plot is
the same in both systems. The details of gettting the plot differ.

Author(s)
Richard M. Heiberger <rmh@temple.edu>

References
Heiberger, Richard M. and Holland, Burt (2004b). Statistical Analysis and Data Display: An Inter-
mediate Course with Examples in S-Plus, R, and SAS. Springer Texts in Statistics. Springer. ISBN
0-387-40270-5.
Heiberger, R. M. and Holland, B. (2006). "Mean–mean multiple comparison displays for families
of linear contrasts." Journal of Computational and Graphical Statistics, 15:937–955.

See Also
mmc, glht in R, multicomp in S-Plus.

axis.i2wt specialized axis function for interaction2wt.

Description
Labels the bottom axis with the x-factor name for each column. Labels the right axis with the
response variable name in all rows.

Usage
axis.i2wt(side, scales, ...)

Arguments
side, scales, ...
See axis.default.

Author(s)
Richard M. Heiberger, with asssistance from Deepayan Sarkar.

See Also
interaction2wt
22 bwplot.formula

bwplot.formula Replacement for bwplot.formula and compute.packet in R to allow

them to work with the new "positioned" class.

Description
Replacement for bwplot.formula and compute.packet in R trellis to allow them to work
with the new "positioned" class.

Usage
## S3 method for class 'formula':
bwplot(x, data = NULL, allow.multiple = is.null(groups) ||
outer, outer = FALSE, auto.key = FALSE, aspect = "fill",
panel = "panel.bwplot", prepanel = NULL, scales = list(),
strip = TRUE, groups = NULL, xlab, xlim, ylab, ylim, box.ratio = 1,
horizontal = NULL,
drop.unused.levels = lattice.getOption("drop.unused.levels"),
...,
default.scales =
if (horizontal)
list(y = list(tck = 0, alternating = FALSE, rot = 0))
else
list(x = list(tck = 0, alternating = FALSE)),
subscripts = !is.null(groups),
subset = TRUE)

compute.packet.new(cond, levels, cond.levels)

Arguments
x, data, allow.multiple, outer, auto.key, aspect, panel, prepanel, scales, strip, g
See bwplot.formula.
cond, levels standard values used by lattice:::compute.packet.
cond.levels

Details
bwplot makes the assumption that the numeric equivalent of a factor f are always integers from
the sequence seq(1:length(levels(f))). When the "positioned" class is introduced,
this is no longer the right assumption. The numeric values need to be chosen from the "position"
attribute.

Value
See bwplot.formula.
ci.plot 23

Author(s)
original by Deepayan Sarkar hDeepayan.Sarkar@R-project.orgi. Modifications by Richard M.
Heiberger <rmh@temple.edu>

## S3 method for class 'lm':

ci.plot(lm.object,
xlim=range(data[, x.name]),
newdata,
conf.level=.95,
data=model.frame(lm.object),
newfit,
ylim=range(newfit$pi.fit),
pch=16,
main.cex=1,
main=list(paste(100*conf.level,
"% confidence and prediction intervals for ",
substitute(lm.object), sep=""), cex=main.cex), ...
)

Arguments
lm.object Linear model for one y and one x variable.
xlim xlim for plot. Default is based on data from which lm.object was con-
structed.
newdata data.frame containing data for which predictions are wanted. The variable
name of the column must be identical to the name of the predictor variable in the
model object. Defaults to a data.frame containing a vector spanning the range
of observed data. User-specified values are appended to the default vector.
conf.level Confidence level for intervals, defaults to .95
24 col.hh

data data extracted from the lm.object

newfit Constructed data.frame containing the predictions,confidence interval, and
prediction interval for the newdata.
ylim ylim for plot. Default is based on the constructed prediction interval.
pch Plotting character for observed points.
main.cex Font size for main title.
main Main title for plot
... Additional arguments to be passed to panel function.

Value
"trellis" object containing the plot.

Note
The predict.lm functions in S-Plus and R differ. The S-Plus function can produce both confi-
dence and prediction intervals with a single call. The R function produces only one of them in a
single call. Therefore the default calculation of newfit within the function depends on the system.

Author(s)
Richard M. Heiberger <rmh@temple.edu>

col.hh Initializing Trellis Displays

Description
Initialization of an R display device with the graphical parameters that rmh prefers.

Usage
col.hh()

Value
List of graphical parameters to be used in the theme argument to the trellis.device or
trellis.par.set functions.
diag.maybe.null 25

Author(s)
Richard M. Heiberger <rmh@temple.edu>

See Also
trellis.device, trellis.par.get

Examples
## function name or result of function call
if.R(r={
trellis.device(theme="col.hh")
trellis.device(theme=col.hh())

## result of function call

trellis.par.set(theme=col.hh())
},s={})

diag.maybe.null Returns a value for the diagonal of NA and NULL arguments.

Description
Returns the argument for the diagonal of NA and NULL arguments. For all other arguments, it calls
the regular diag function.

Usage
diag.maybe.null(x, ...)

Arguments
x matrix, vector, NA,
... Other arguments to diag.

Author(s)
Richard M. Heiberger (rmh@temple.edu)

export.eps Exports a graph to an EPS file.

Description

Exports a graph from the current device in R, or the graphsheet in S-Plus, to an EPS file.

Usage

export.eps(FileName.in, Name.in="GSD2", ...)

Arguments

FileName.in name of file to be created.

Name.in Name of graphsheet in S-Plus, ignored in R.
... other arguments in R, ignored in S-Plus.

Author(s)

Richard M. Heiberger <rmh@temple.edu>

See Also

dev.copy2eps in R, export.graph in S.

Examples
## Not run:
trellis.device()
plot(1:10)
export.eps("abcd.eps")
## End(Not run)

extra Miscellaneous functions that I wish were in or consistent between S-

Plus and R.

Description

Miscellaneous functions that I wish were in or consistent between S-Plus and R.

28 gof.calculation

Usage
as.rts(x, ...)

## S3 method for class 'ts':

units(x)

title.trellis(main = NULL, sub = NULL, xlab = NULL, ylab = NULL,

line = NA, outer = FALSE, axes=NULL, ...)

title.grob(main=NULL, y=.985, gp=gpar(cex=1.5))

as.character.arima.model(x, model=arima.model(x), ...)

arima.model(x)

coef.arima.HH(object, ...)

.arima.info.names.not.ordered (model)

Arguments
x vector or time series
... Additional arguments.
main, sub, xlab, ylab, line, outer, axes
See title.
model A time series model specification in the S-Plus notation.
object "arima" object in S-Plus.
y, gp See grid.text in R.

Author(s)
Richard M. Heiberger (rmh@temple.edu)

gof.calculation Calculate Box–Ljung Goodness of Fit for ARIMA models in S-Plus.

Description
Calculate Box–Ljung Goodness of Fit for ARIMA models in S-Plus. In R we use the Box.test
function.
grid.yaxis.hh 29

Usage
gof.calculation(acf.list, gof.lag, n, n.parms)

Arguments
acf.list An "acf" object.
gof.lag The number of model parameters is the number of lags to use for computing the
Portmanteau goodness of fit statistic
n Number of residuals in model.
n.parms Number of AR and MA parameters in the model.

Details
This function is isolated from the S-Plus arima.diag function. It is used only in S-Plus.

Value
See the gof value described in arima.diag in S-Plus.

Author(s)
Richard M. Heiberger (rmh@temple.edu)

grid.yaxis.hh make x- and y-axis labels

Description
uses modified older version of grid functions. Includes optional specification of the axis labels.
30 hh

Usage
grid.yaxis.hh(at = NULL, label = TRUE, main = TRUE, gp = gpar(),
draw = TRUE, vp = NULL, labels)

make.yaxis.hh.labels(at, main, labels = at)

grid.xaxis.hh(at = NULL, label = TRUE, main = TRUE, gp = gpar(),

draw = TRUE, vp = NULL, labels)

make.xaxis.hh.labels(at, main, labels = at)

Arguments
at, label, main, gp, draw, vp
See link[grid]{grid.xaxis}.
labels label values if you don’t want the defaults

Value
See link[grid]{grid.xaxis}.

Author(s)
Richard M. Heiberger <rmh@temple.edu>

hh Resolve filenames relative to the HH directory.

Description
The pathnames for all the datasets and executable files in the online files accompanying Heiberger
and Holland (2004) are given relative to the options()$HH.ROOT.DIR directory.

Usage
hh(file)

Arguments
file Character string giving the pathname of a file in the HH online files relative to the
options()$HH.ROOT.DIR directory. file is the name exactly as specified
in Heiberger and Holland (2004). The hh function resolves those names to the
location where the files are stored on your computer.
hov 31

Details

The datasets and code listings in Heiberger and Holland (2004) are all given paths relative to the
beginning of the hh directory, which must be stored as options()$HH.ROOT.DIR
If you installed the HH library from an R or S-Plus package, then option options()$HH.ROOT.DIR
is set automatically.
If you installed the HH library by unzipping the downloaded online files (not a package), then
you must manually define the HH.ROOT.DIR option in your working directory (for example,
"c:/HOME/yourname/.Data" in S-Plus, or the .GlobalEnv in R). You may do this in your
.First function in either system. Or you may enter at the command prompt a statement of the
form options(HH.ROOT.DIR="c:/HOME/hh") in Windows or options(HH.ROOT.DIR="/usr/users/hh")
in Unix. See Appendix B of Heiberger and Holland (2004) for further details. The options state-
ment may need to be modified to match the location of the hh directory on your machine. If you
use more than one computer, you may need a different value for the HH.ROOT.DIR option on
each machine. This is the only change you will need to make in order to run any of our software or
examples. The hh function itself is unchanged.

Value

Fully expanded, absolute pathname for the input filename.

Author(s)

Richard M. Heiberger <rmh@temple.edu>

References

Heiberger, Richard M. and Holland, Burt (2004b). Statistical Analysis and Data Display: An Inter-
mediate Course with Examples in S-Plus, R, and SAS. Springer Texts in Statistics. Springer. ISBN
0-387-40270-5.

Examples
hotdog <- read.table(hh("datasets/hotdog.dat"), header=TRUE)

hov Homogeneity of Variance

Description

Oneway analysis of variance makes the assumption that the variances of the groups are equal.
Brown and Forsyth, 1974 present the recommended test of this assumption. The Brown and Forsyth
test statistic is the F statistic resulting from an ordinary one-way analysis of variance on the absolute
deviations from the median.
32 hov

Usage
hov(x, data = sys.parent(), method = "bf") ## x is a formula

## users will normally use the formula above and will not call the
## method directly.
hov.bf(x, group, ## x is the response variable
y.name = deparse(substitute(x)),
group.name = deparse(substitute(group)))

Arguments
x Formula appropriate for oneway anova in hov. Response variable in hov.bf.
data data.frame
method Character string defining method. At this time the only recognized method is
"bf" for the Brown–Forsyth method.
group factor.
y.name name of response variable, defaults to variable name in formula.
group.name name of factor, defaults to variable name in formula.

Value
"htest" object for the hov test.

Author(s)
Richard M. Heiberger <rmh@temple.edu>

References
Heiberger, Richard M. and Holland, Burt (2004b). Statistical Analysis and Data Display: An Inter-
mediate Course with Examples in S-Plus, R, and SAS. Springer Texts in Statistics. Springer. ISBN
0-387-40270-5.
Brown, M. B. and Forsyth, A. B. (1974). Robust tests for equality of variances. Journal of the
American Statistical Association, 69:364–367.

hov(wt.gain ~ diet, data=turkey)

plot.hov(wt.gain ~ diet, data=turkey)
if.R 33

if.R Conditional Execution for R or S-Plus

Description

if.R uses the is.R function to determine whether to execute the expression in the r argument or
the expression in the s argument.

Usage

if.R(r, s)

Arguments

r Any R expression, including a group of expressions nested in braces. Assign-

ments made in this expression are available to the enclosing function.
s Any S-Plus expression, including a group of expressions nested in braces. As-
signments made in this expression are available to the enclosing function.

Details

Not all functions are in both implementations of the S language. In particular, panel functions for
lattice in R (based on grid graphics) are very different from panel functions for trellis
(based on the older graphics technology) in S-Plus.

Value

The result of the executed expression.

Author(s)

Richard M. Heiberger <rmh@temple.edu>

See Also

is.R

Examples
if.R(r={"This is R."},
s={"This is S-Plus"})
34 interaction.positioned

interaction.positioned
interaction method for positioned factors.

Description
This is intended to be a method for interaction for positioned factors. Since interaction is
not currently implemented as a generic, interaction.positioned is a standalone function.
The result is assigned a position. The position for each interaction level is the position of the
corresponding a factor plus a scaled level of the b factor. The default scale is .1.

Usage
interaction.positioned(..., ## exactly two factors
drop = FALSE, sep = ".",
b.offset=0,
b.scale=.1)

Arguments
... exactly two factors. The first factor a is used as the major factor in sort order.
The second factor b is used as minor factor in sort order.
b.offset amount added to position(b) to adjust appearance.
b.scale scale to relate units of position(a) to units of position(b).
drop, sep See factor.

Value
"positioned" object containing the ordinary interaction with a "position" attribute.

Author(s)
Richard M. Heiberger <rmh@temple.edu>

## S3 method for class 'formula':

interaction2wt(x, data = sys.parent(), responselab, ...)

## Default S3 method:
interaction2wt(x,
response.var,
responselab = deparse(substitute(y)),
relation = list(x = "same", y = "same"),
x.relation = relation$x,
y.relation = relation$y,
digits = 3,
x.between=if (label.as.interaction.formula) 0 else 1,
y.between=if (label.as.interaction.formula) 0 else 1,
between,
cex = 0.75,
rot=c(0,0),
panel.input = panel.interaction2wt,
strip.input =
if (label.as.interaction.formula) strip.default
else strip.interaction2wt,
par.strip.text.input = list(cex=.7, responselab=responselab),
scales.additional,
main.in =
paste(responselab,
": main effects and 2-way interactions",
sep = ""),
xlab = list(labels = ""),
ylab = list(labels = ""),
simple=FALSE,
box.ratio=if (simple) .32 else 1,
label.as.interaction.formula=TRUE,
...,
main.cex)
36 interaction2wt

Arguments
Arguments when x is a formula.
x The object on which method dispatch is carried out.
For the "formula" method, a formula describing the response variable and
factors. The formula is generally of the form y ~ g1 + g2 + .... There
may be one or more factors in the formula.
For the "default" method, data.frame of factors. This is usually con-
structed by formula method from the input data and the input formula.
data For the formula method, a data frame containing values for any variables in
the formula. In the R version, if not found in data, or if data is unspecified,
the variables are looked for in the environment of the formula.
responselab Character name of response variable, defaults to the name of the response vari-
able in the formula.
... additional arguments, primarily trellis arguments.
response.var For the "default" method, the response variable. This is usually constructed
by formula method from the input data and the input formula.
simple logical. TRUE if simple effects are to be displayed. Arguments simple.offset,
simple.scale, and col.by.row may also be needed. See panel.interaction2wt
for details.
box.ratio bwplot in R or bwplot in S-Plus.
relation trellis argument.
x.relation x value of relation argument.
y.relation y value of relation argument.
digits doesn’t do anything at the moment
x.between x value of between argument.
y.between y value of between argument.
between trellis/lattice between argument. If used, between has precedence over both
the x.between and y.between arguments.
cex S-Plus: changes the size of the median dot in the boxplots. R: doesn’t do any-
thing.
panel.input panel function. Default is panel.interaction2wt.
label.as.interaction.formula
logical. If TRUE, each panel has a single strip label of the form y ~ a | b. If
FALSE, each panel has a pair of strip labels, one for the trace factor and one
for the x factor.
strip.input strip function. Default depends on the value of label.as.interaction.formula.
par.strip.text.input
par.strip.text argument.
scales.additional
additional arguments to scales argument.
main.in Text of main title.
interaction2wt 37

xlab No effect.
ylab No effect.
main.cex cex for main title.
rot Rotation of x tick labels and y tick labels. Only 0 and 90 will look good.

Value
"trellis" object containing the plot.

Author(s)
Richard M. Heiberger <rmh@temple.edu>

position(esoph$alcgp) <- 2:5

position(esoph$tobgp) <- 2:5

print(
interaction2wt(rate ~ agegp + alcgp + tobgp, esoph, rot=c(90,0))
)
esoph.aov <- aov(rate ~ agegp + alcgp + tobgp, data=esoph)
print(
anova(esoph.aov)
)
}, s={})

intxplot Interaction plot, with an option to print standard error bars.

Description
Interaction plot, with an option to print standard error bars. There is an option to offset group lines
to prevent the bars from overprinting.

Usage
intxplot(x, data=sys.parent[1], groups.in,
scales,
key.length=1,
key.lines,
key=TRUE,
trace.factor.name=deparse(substitute(groups.in)),
x.factor.name=x.factor,
xlab=x.factor.name,
main=list(main.title, cex=main.cex),
condition.name="condition",
panel="panel.intxplot",
summary.function="sufficient",
se,
...,
data.is.summary=FALSE,
main.title=paste(
intxplot 39

"Interactions of", trace.factor.name, "and",

x.factor.name,
if (length(x[[3]]) > 1)
paste("|", condition.name.to.use)),
main.cex=1.5)

panel.intxplot(x, y, subscripts, groups, type = "l", ..., se, cv=1.96,

offset.use=(!missing(groups) && !missing(se)),
offset.scale=2*max(as.numeric(groups)),
offset=
as.numeric(groups[match(levels(groups), groups)]) / offset.scale,
rug.use=offset.use)

Arguments
x For intxplot, a formula with a factor as the predictor variable. For panel.intxplot,
standard argument for panel functions.
data data.frame, as used in xyplot.
groups.in groups.in, as used in xyplot.
scales Optional, additional arguments for the standard scales in xyplot.
key.length Number of columns in the key.
key.lines default value for the lines argument of key.
key logical. If TRUE, draw the key.
trace.factor.name
Name of the grouping variable.
x.factor.name
name of the dependent variable.
xlab as in xyplot, defaults to the name of the predictor variable from the formula.
main as in xyplot. Defaults to the main.title argument.
panel as in xyplot. Defaults to the "panel.intxplot".
condition.name
name of the conditioning variable.
summary.function
The default sufficient finds the mean, standard deviation, and sample size
of the response variable for each level of the conditioning factor. See sufficient.
se standard errors to be passed to panel.intxplot. se Missing, logical, or
a numeric vector. If missing or FALSE, standard errors are not plotted. If
se=TRUE in intxplot, the standard errors are calculated from the suffi-
cient statistics for each group as the group’s standard deviation divided by the
square root of the group’s observation count. If se is numeric vector, it is
evaluated in the environment of the sufficient statistics. the se argument to
panel.intxplot must be numeric.
... In intxplot, arguments for panel.intxplot. In panel.intxplot,
arguments for panel.superpose.
40 intxplot

data.is.summary
logical, defaults to FALSE under the assumption that the input data.frame is
the original data and the intxplot function will generate the summary infor-
mation (primarily standard deviation sd and number of observations nobs for
each group). When TRUE, the standard error calculation assumes variables sd
and nobs are in the dataset.
main.title Default main title for plot.
main.cex Default character expansion for main title.
y, subscripts, groups, type
Standard arguments for panel functions.
cv critical value for confidence intervals. Defaults to 1.96.
offset.use logical. If TRUE, offset the endpoints of each group.
offset.scale Scale number indicating how far apart the ends of the groups will be placed.
Larger numbers make them closer together.
offset Actual numbers by which the end of the groups are offset from their nominal
location which is the as.numeric of the group levels.
rug.use logical. If TRUE, display a rug for the endpoints of each group.

Value
"trellis" object.

Author(s)
Richard M. Heiberger <rmh@temple.edu>

rhiz.clover <- read.table(hh("datasets/rhiz3-clover.dat"), header=TRUE)

rhiz.clover$comb <- factor(rhiz.clover$comb,
labels=c("clover","clover+alfalfa"))
position(rhiz.clover$comb) <- c(2,5)
rhiz.clover$strain <-
factor(rhiz.clover$strain,
labels=c('3DOk1','3DOk5','3DOk4','3DOk7','3DOk13','k.comp'))
rhiz.clover$Npg <- rhiz.clover$nitro / rhiz.clover$weight

## interaction plot, no se
intxplot(Npg ~ strain, groups=comb, data=rhiz.clover)

## interaction plot, individual se for each treatment combination

intxplot(Npg ~ strain, groups=comb, data=rhiz.clover, se=TRUE)
ladder 41

## Rescaled to allow the CI bars to stay within the plot region

intxplot(Npg ~ strain, groups=comb, data=rhiz.clover, se=TRUE,
ylim=range(rhiz.clover$Npg))

## interaction plot, common se based on ANOVA table

intxplot(Npg ~ strain, groups=comb, data=rhiz.clover,
se=sqrt(sum((nobs-1)*sd^2)/(sum(nobs-1)))/sqrt(5))

## Rescaled to allow the CI bars to stay within the plot region

intxplot(Npg ~ strain, groups=comb, data=rhiz.clover,
se=sqrt(sum((nobs-1)*sd^2)/(sum(nobs-1)))/sqrt(5),
ylim=range(rhiz.clover$Npg))

## change distance between endpoints

intxplot(Npg ~ strain, groups=comb, data=rhiz.clover,
se=TRUE, offset.scale=20)

## When data includes the nobs and sd variables, data.is.summary=TRUE is needed.

intxplot(Npg ~ strain, groups=comb,
se=sqrt(sum((nobs-1)*sd^2)/(sum(nobs-1)))/sqrt(5),
data=sufficient(rhiz.clover, y="Npg", c("strain","comb")),
data.is.summary=TRUE,
ylim=range(rhiz.clover$Npg))

ladder Draw a "ladder of powers" plot, plotting each of several powers of y

against the same powers of x.

Description
Draw a "ladder of powers" plot, plotting each of several powers of y against the same powers of x.
The powers are
result <- data.frame(-1/x, -1/sqrt(x), log(x), sqrt(x), x, x^2)
names(result) <- c(-1, -.5, 0, .5, 1, 2)

Usage
ladder(formula.in, data=sys.parent(),
main.in="Ladders of Powers",
panel.in=panel.cartesian,
xlab=deparse(formula.in[[3]]),
ylab=deparse(formula.in[[2]]),
scales=list(alternating=if.R(s=TRUE, r=FALSE),
labels=FALSE, ticks=FALSE, cex=.6),
par.strip.text=list(cex=.6),
cex=.5, pch=16, between=list(x=.3, y=.3),
dsx=xlab,
dsy=ylab,
42 ladder

ladder.function=ladder.f,
strip.number=1,
strip.names,
strip.style=1,
strip,
oma=c(0,0,0,0), ## S-Plus
axis3.line=.61,
layout=c(length(tmp$x.power), length(tmp$y.power)),
axis.key.padding = 10, ## R right axis
key.axis.padding = 10, ## R top axis
...)

ladder3(x, y,
dsx=deparse(substitute(x)),
dsy=deparse(substitute(y)),
ladder.function=ladder.f)

ladder.f(x)

ladder.fstar(x)

strip.ladder(which.given,
which.panel,
var.name,
factor.levels,
shingle.intervals,
par.strip.text=trellis.par.get("add.text"),
strip.names=c(TRUE,TRUE),
style=1,
...)

Arguments
formula.in A formula with exactly one variable on each side.
data data.frame
main.in main title for xyplot
panel.in panel.cartesian has many arguments in addition to the arguments in panel.xyplot.
Any replacement panel function must have those argument names, even if it
doesn’t do anything with them.
xlab, ylab Trellis arguments, default to right- and left-sides of the formula.in.
strip Strip function. Our default is strip.ladder (see below). The other viable
argument value is FALSE.
cex, pch, between, scales, layout
arguments for xyplot.
dsx, dsy Names to be used as level names in ladder.function for the generated fac-
tor distinguishing the powers. They default to xlab, ylab. For long variable
ladder 43

names, an abbreviated name here will decrease clutter in the ladder of powers
plot. These names are not visible in the plot when strip=FALSE.
ladder.function
function to use to create data.frame of powers of input variable.
strip.number Number of strip labels in each panel of the display. 0: no strip labels; 1: one
strip label of the form y^p ~ x^q; 2: two strip labels of the form ylab:
y^p and xlab: x^q, where p and q are the powers returned by ladders;
y and x are the arguments dsy and dsx.
strip.style style argument to strip.
oma argument to par in S-Plus.
... other arguments to xyplot.
axis3.line extra space to make the top axis align with the top of the top row of panels. Trial
and error to choose a good value.
axis.key.padding
Extra space on right of set of panels in R.
key.axis.padding
Extra space on top of set of panels in R.
x, y variables.
which.given, which.panel, var.name, factor.levels, shingle.intervals, par.strip.tex
See strip.default in R or strip.default in S-Plus.
strip.names, style
We always print the strip.names in style=1. Multicolored styles are too
busy.

Details
The ladder function uses panel.cartesian which is defined differently in R (using grid
graphics) and S-Plus (using traditional graphics). Therefore the fine control over appearance uses
different arguments or different values for the same arguments.

Value
ladder returns a "trellis" object.
The functions ladder.fstar and ladder.f take an input vector x of non-negative values
and construct a data.frame by taking the input to the powers c(-1, -.5, 0, .5, 1, 2),
one column per power. ladder.f uses the simple powers and ladder.fstar uses the scaled
Box–Cox transformation.

ladder.fstar ladder.fstar notation

(x^p - 1)/p (x^p - 1)/p p
(1/x - 1)/(-1) (1/x - 1)/(-1) -1.0
(1/sqrt(x)-1)/(-.5) (1/sqrt(x)-1)/(-.5) -0.5
log(x) log(x) 0.0
((sqrt(x)-1)/.5) ((sqrt(x)-1)/.5) 0.5
x-1 x-1 1.0
(x^2 - 1)/2 (x^2 - 1)/2 2.0
44 ladder

ladder3 takes two vectors as arguments. It returns a data.frame with five columns:
X, Y data to be plotted. The column X contains the data from the input x taken to all the powers
and aligned with the similarly expanded column Y.
x, y symbolic labeling of the power corresponding to X,Y.
group result from pasting the labels in x, y with * between them.

Author(s)
Richard M. Heiberger <rmh@temple.edu>

References
Heiberger, Richard M. and Holland, Burt (2004b). Statistical Analysis and Data Display: An Inter-
mediate Course with Examples in S-Plus, R, and SAS. Springer Texts in Statistics. Springer. ISBN
0-387-40270-5.
Hoaglin, D. C., Mosteller, F., and Tukey, J. W., editors (1983). Understanding Robust and Ex-
ploratory Data Analysis. Wiley.
Box, G. E. P. and Cox, D. R. (1964). An analysis of transformations. J. Royal Statist Soc B,
26:211–252.

## Default: single strip label per panel

ladder(life.exp ~ ppl.per.phys, data=tv,
main="Ladder of Powers for Life Expectancy and People per Physician",
dsx="ppp", dsy="le")

## double strip label

if.R(r=
ladder(life.exp ~ ppl.per.phys, data=tv,
main="Ladder of Powers for Life Expectancy and People per Physician",
legendGrob2wt 45

strip.number=2,
dsx="ppp", dsy="le")
,s=
ladder(life.exp ~ ppl.per.phys, data=tv,
main="Ladder of Powers for Life Expectancy and People per Physician",
strip.number=2,
dsx="ppp", dsy="le",
axis3.line=1.2)
)

## turn off strip labels

if.R(r=
ladder(life.exp ~ ppl.per.phys, data=tv,
strip.number=0,,
main="Ladder of Powers for Life Expectancy and People per Physician")
,s=
ladder(life.exp ~ ppl.per.phys, data=tv,
strip.number=0,,
main="Ladder of Powers for Life Expectancy and People per Physician",
axis3.line=0)
)

legendGrob2wt place separate keys to the left of each row of a trellis

Description
Each key is created and then inserted into a single grob.

Usage
legendGrob2wt(...)

Arguments
... key1, key2, etc. Each key will normally be the result of a draw.key with
draw=FALSE.

Value
A Grid frame object (that inherits from ’grob’).

Author(s)
Richard M. Heiberger, with asssistance from Deepayan Sarkar.

lm.case case statistics for regression analysis

Description
Case statistics for regression analysis. lm.case calculates the statistics. plot.case plots the
cases, one statistic per panel, and illustrates and itemizes all observations for which the stan-
dard thresholds are exceeded. plot.case returns a "trellis" object containing the plot
and also places the row.names of the flagged observations in the variable .lm.case.large.
panel.case is a panel function for plot.case.

Usage
lm.case(fit, lms = summary.lm(fit), lmi = lm.influence(fit))

plot.case(x, fit,
which=c("stu.res","si","h","cook","dffits",
dimnames(x)[[2]][-(1:8)]), ##DFBETAS
between.in=list(y=4, x=9),
oma=c(0,0,0,4), cex.threshold=if.R(r=1.2, s=1.6),
main.in=list(
paste(deparse(fit$call), collapse=""),
cex=main.cex),
sigma.in=summary.lm(fit)$sigma,
p.in=summary.lm(fit)$df[1]-1,
obs.large=".lm.case.large",
obs.large.env=if.R(r=globalenv(), s=0),
main.cex=NULL,
...)

panel.case(x, y, subscripts, rownames, group.names,

nn, pp, ss, cex.threshold,
par.settings, ## R only. S-Plus ignores this argument
obs.large, obs.large.env,
...)

Arguments
fit "lm" object computed with x=TRUE
lms summary.lm(fit)
lmi lm.influence(fit)
x In plot.case, the matrix output from lm.case containing case diagnostics
on each observation in the original dataset. In panel.case, the x variable to
be plotted
which In plot.case, the names of the columns of x that are to be graphed.
lm.case 47

between.in between trellis/lattice argument.

oma In S-Plus, change par()$oma to make room for the threshold values. A warn-
ing is printed when par()$oma is changed as the delayed printing of trellis
objects implies we can’t return it to the original value automatically. In R,
this argument is ignored. Instead, we use the par.settings argument to
xyplot inside plot.case. The par.settings becomes one component
of the "trellis" object that is the value of plot.case and is therefore
automatically applied every time the object is printed.
cex.threshold
Multiplier for cex for the threshold values.
main.in main title for xyplot. The default main title displays the linear model formula
from fit.
sigma.in standard error for the fit.
p.in The number of degrees of freedom associated with the fitted model.
obs.large Object name where the names of all observations for which the standard thresh-
olds are exceeded will be stored. The default name is .lm.case.large.
obs.large.env
Frame in S-Plus (defaults to 0) and environment in R (defaults to globalenv())
where obs.large will be stored.
main.cex cex for main title.
... other arguments to xyplot
y the y variable to be plotted.
nn number of rows in original dataset.
pp The number of degrees of freedom associated with the fitted model.
ss standard error for the fit.
subscripts trellis/lattice argument, position in the reshaped dataset constructed by plot.case
before calling xyplot.
rownames row name in the original data.frame.
group.names names of the individual statistics.
par.settings Used in R as part of the call to xyplot. Although this argument is not used in
the panel function, it is needed as a formal argument in S-Plus to absorb it out
of ... and thereby prevent it from being forwarded to points.

Details
lm.influence is part of S-Plus and R lm.case and plot.case are based on: Section 4.3.3
"Influence of Individual Obervations in Chambers and Hastie", Statistical Models in S.

Value
lm.case returns a matrix, with one row for each observation in the original dataset. The columns
contain the diagnostic statistics: e (residuals), h* (hat diagonals), si* (deleted standard deviation),
sta.res (standardized residuals), stu.res* (Studentized deleted resididuals), dffit (differ-
ence in fits, change in predicted y when observation i is deleted), dffits* (standardized difference
48 lm.case

in fits, standardized change in predicted y when observation i is deleted), cook* (Cook’s distance),
and DFBETAs* (standardized difference in regression coefficients when observation i is deleted,
one for each column of the x-matrix, including the intercept).
plot.case returns a "trellis" object containing the plot (including the starred columns by
default) and also places the row.names of the flagged observations in the variable .lm.case.large.
The variable .lm.case.large is placed by default into frame 0 in S-Plus and into globalenv()
in R.
panel.case is a panel function for plot.case. The variable .lm.case.large is created
one column at a time inside the panel function.

Author(s)

Richard M. Heiberger <rmh@temple.edu>

References

See Also

lm.influence in R, lm.influence in S-Plus.

Examples

kidney <- read.table(hh("datasets/kidney.dat"), header=TRUE)

kidney2.lm <- lm(clearance ~ concent + age + weight + concent*age,

data=kidney, x=TRUE) ## the lm object must be computed with x=TRUE

kidney2.case <- lm.case(kidney2.lm)

## this picture looks much better in portrait, specification is device dependent

## trellis.device(postscript, horizontal=TRUE) ## postscript
## trellis.device(orientation="portrait") ## S-Plus graphsheet

plot.case(kidney2.case, kidney2.lm, par.strip.text=list(cex=.9),

layout=c(2,3))

.lm.case.large ## file placed by default into frame 0 in S-Plus

## and into globalenv() in R
mcalinfct 49

mcalinfct MCA multiple comparisons analysis (pairwise)

Description
MCA multiple comparisons analysis (pairwise). We calculate the contrast matrix for all pairwise
comparisons, taking account of covariates and interactions.

Usage
mcalinfct(model, focus,
mmm.data=model$model,
formula.in=terms(model),
linfct.Means=
multcomp:::meanslinfct(model, focus, mmm.data, formula.in),
type="Tukey"
)

Arguments
model aov object
focus name of one of the factors in the model, as a character object.
mmm.data data.frame from which the model was estimated. Normally, the default is
the correct value.
formula.in formula of the model which was estimated. Normally, the default is the cor-
rect value. The use of the terms function honors the keep.order=TRUE if
it was specified.
linfct.Means Contrast matrix for the adjusted means of each level of the focus factor. Nor-
mally, the default is the correct value.
type Name of the multiple comparison procedure to be used. See contrMat.

Value
Matrix to be used as a value for the linfct argument to glht.

Note
This function provides results similar to the mcp(focusname="Tukey") argument to glht. I
think it provides better values for covariate and interaction terms.

Author(s)
Richard M. Heiberger <rmh@temple.edu>

mmc MMC (mean–mean multiple comparisons) plots.

Description
Constructs a "mmc.multicomp" object from the formula and other arguments. The object must
be explicitly plotted.

Usage
glht.mmc(model, ...) ## R

## S3 method for class 'glht':

glht.mmc(model, ...)

## S3 method for class 'lm':

glht.mmc(model, ## lm object
linfct=NULL,
focus=
if (is.null(linfct))
{
if (length(model$contrasts)==1) names(model$contrasts)
else stop("focus or linfct must be specified.")
}
else
{
if (is.null(names(linfct)))
stop("focus must be specified.")
else names(linfct)
},
ylabel=as.character(terms(model)[[2]]),
lmat=t(linfct),
lmat.rows=-1,
lmat.scale.abs2=TRUE,
estimate.sign=1,
order.contrasts=TRUE,
level=.95,
calpha=NULL,
alternative = c("two.sided", "less", "greater"),
...
)

multicomp.mmc(..., comparisons="mca", ## S-Plus

mmc 51

lmat, lmat.rows=-1,
lmat.scale.abs2=TRUE,
ry,
plot=TRUE,
crit.point,
iso.name=TRUE,
estimate.sign=1,
x.offset=0,
order.contrasts=TRUE,
main,
main2)

## S3 method for class 'mmc.multicomp':

x[..., drop = TRUE]

Arguments
model "aov" object in "lm" method.
ylabel name of the response variable.
lmat contrast matrix as in the S-Plus multicomp. The convention for lmat in R is
to use the transpose of the linfct component produced by glht. Required
for user-specified contrasts.
lmat.rows rows in lmat for the focus factor.
focus define the factor to compute contrasts of. See mcp in R.
linfct In R, see glht.
... other arguments. alternative and base are frequently used with glht.
comparisons argument to multicomp
lmat.scale.abs2
logical, scale the contrasts in the columns of lmat to make the sum of the
absolute values of each column equal 2.
estimate.sign
numeric. If 0, leave contrasts in the default lexicographic direction. If positive,
force all contrasts to positive, reversing their names if needed (if contrast A-B is
negative, reverse it to B-A). If negative, the force all contrasts to positive.
order.contrasts
sort the contrasts in the (mca, none, lmat) components by height on the MMC
plot. This will place the contrasts in the multicomp plots in the same order as in
the MMC plot.
alternative Direction of alternative hypothesis. See confint in R. S-Plus multicomp
uses the argument bounds for this concept.
level Confidence level. Defaults to 0.95.
crit.point, calpha
critical value for the tests. The value from the specified multicomp method
is used for the user-specified contrasts when lmat is specified. This argument
is called crit.point with multicomp in S-Plus and calpha when used
52 mmc

with glht and confint in R. In R, with a large number of levels for the
focus factor, calpha should be specified. See notes below for discussion of
the timing issues and the examples for an illustration how to use calpha.
plot logical, display the plot if TRUE.
ry, iso.name, x.offset, main, main2
arguments to plot.mmc.multicomp.
x, drop See "[".

Details
By default, if lmat is not specified, we plot the isomeans grid and the pairwise comparisons for
the focus factor. By default, we plot the specified contrasts if the lmat is specified.
We get the right contrasts automatically if the aov is oneway. If we specify an lmat for oneway it
must have a leading row of 0.
For any more complex design, we must study the lmat from the mca component of the result to
see how to construct the lmat (with the extra rows as needed) and how to specify the lmat.rows
corresponding to the rows for the focus factor.
glht.mmc in R works from either an "glht" object or an "aov" object. multicomp.mmc in
S-Plus works from an "aov" object.

Value
An "mmc.multicomp" object contains either the first two or all three of the "multicomp"
components
mca
none
lmat
mca Object containing the pairwise comparisons.
none Object comparing each mean to 0.
lmat Object for the contrasts specified in the lmat argument.
"[.mmc.multicomp" is a subscript method.

Note
The multiple comparisons calculations in R and S-Plus use completely different functions. MMC
plots in R are constructed by glht.mmc based on glht. MMC plots in S-Plus are constructed by
multicomp.mmc based on the S-Plus multicomp. The MMC plot is the same in both systems.
The details of getting the plot differ.
Function glht.mmc calls glht and confint.glht. With a large number of levels for the focus
factor, the confint function is exceedingly slow (80 minutes for 30 levels on 1.5GHz Windows
XP). Therefore, always specify calpha to reduce the time to under a second for the same example.
plot.mmc.multicomp chooses sensible defaults for its many arguments. They will often need
manual adjustment. The examples show several types of adjustments. We have changed the center-
ing and scaling to avoid overprinting of label information. By default the significant contrasts are
shown in a more intense color than the nonsignificant contrasts. We have an option to reduce the
color intensity of the isomeans grid.
mmc 53

Author(s)
Richard M. Heiberger <rmh@temple.edu>

See Also
as.multicomp, plot.mmc.multicomp

Examples
## Use glht.mmc with R.
## Use multicomp.mmc with S-Plus.

## data and ANOVA

catalystm <- read.table(hh("datasets/catalystm.dat"), header=FALSE,
col.names=c("catalyst","concent"))
catalystm$catalyst <- factor(catalystm$catalyst, labels=c("A","B","C","D"))

if.R(r=
bwplot(concent ~ catalyst, data=catalystm,
scales=list(cex=1.5),
ylab=list("concentration", cex=1.5),
xlab=list("catalyst",cex=1.5))
,s=
t(bwplot(catalyst ~ concent, data=catalystm,
scales=list(cex=1.5),
xlab=list("concentration", cex=1.5),
ylab=list("catalyst",cex=1.5)))
)

catalystm1.aov <- aov(concent ~ catalyst, data=catalystm)

summary(catalystm1.aov)

catalystm.mca <-
if.R(r=glht(catalystm1.aov, linfct = mcp(catalyst = "Tukey")),
s=multicomp(catalystm1.aov, plot=FALSE))
plot(catalystm.mca) ## S-Plus always and R for small number of levels
if.R(s={},
r=plot(confint(catalystm.mca, calpha=qtukey(.95, 4, 12)/sqrt(2))))
## R for large number of levels
catalystm.mca
54 mmc

## pairwise comparisons
catalystm.mmc <-
if.R(r=glht.mmc(catalystm1.aov, linfct = mcp(catalyst = "Tukey")),
s=multicomp.mmc(catalystm1.aov, plot=FALSE))
catalystm.mmc
plot(catalystm.mmc)
plot(catalystm.mmc$mca)
plot(catalystm.mmc$none)

if.R(s={},
r={## R for large number of levels
plot(confint(catalystm.mmc$mca$glht,
calpha=catalystm.mmc$mca$crit.point))
plot(confint(catalystm.mmc$none$glht,
calpha=catalystm.mmc$none$crit.point))
}
)

## user-specified contrasts
catalystm.lmat <- cbind("AB-D" =c(0, 1, 1, 0,-2),
"A-B" =c(0, 1,-1, 0, 0),
"ABD-C"=c(0, 1, 1,-3, 1))
if.R(r=catalystm.lmat <- catalystm.lmat[-2,],
s={})
dimnames(catalystm.lmat)[[1]] <- dimnames(catalystm.mmc$mca$lmat)[[1]]
zapsmall(catalystm.lmat)
if.R(s=dimnames(catalystm.mca$lmat)[[1]],
r=dimnames(catalystm.mca$linfct)[[2]])

catalystm.mmc <-
if.R(r=glht.mmc(catalystm1.aov, linfct = mcp(catalyst = "Tukey"),
lmat=catalystm.lmat)
,s=multicomp.mmc(catalystm1.aov, lmat=catalystm.lmat,
plot=FALSE)
)

catalystm.mmc
plot(catalystm.mmc)

plot(catalystm.mmc$mca)
plot(catalystm.mmc$none)
plot(catalystm.mmc$lmat)

## Dunnett's test
weightloss <- read.table(hh("datasets/weightloss.dat"), header=TRUE)
weightloss <- data.frame(loss=unlist(weightloss),
group=rep(names(weightloss), rep(10,5)))
if.R(r=
bwplot(loss ~ group, data=weightloss,
scales=list(cex=1.5),
ylab=list("Weight Loss", cex=1.5),
mmc 55

xlab=list("group",cex=1.5))
,s=
t(bwplot(group ~ loss, data=weightloss,
scales=list(cex=1.5),
xlab=list("Weight Loss", cex=1.5),
ylab=list("group",cex=1.5)))
)

weightloss.aov <- aov(loss ~ group, data=weightloss)

summary(weightloss.aov)

if.R(r={
group.count <- table(weightloss$group)
},s={})

tmp.dunnett <-
if.R(r=
glht(weightloss.aov,
linfct=mcp(group=contrMat(group.count, base=4)),
alternative="greater")
,s=
multicomp(weightloss.aov,
method="dunnett", comparisons="mcc",
bounds="lower", control=4,
valid.check=FALSE)
)
plot(tmp.dunnett)

tmp.dunnett.mmc <-
if.R(r=
glht.mmc(weightloss.aov,
linfct=mcp(group=contrMat(group.count, base=4)),
alternative="greater")
,s=
multicomp.mmc(weightloss.aov,
method="dunnett", comparisons="mcc",
bounds="lower", control=4,
valid.check=FALSE, plot=FALSE)
)

tmp.dunnett.mmc
plot(tmp.dunnett.mmc)

## two-way ANOVA
display <- read.table(hh("datasets/display.dat"), header=TRUE)
display$panel <- factor(display$panel) ## display$panel <- positioned(display$panel, value=
display$emergenc <- factor(display$emergenc)

displayf.aov <- aov(time ~ emergenc * panel, data=display)

anova(displayf.aov)

## multiple comparisons
56 mmc

tmp <- if.R(

r=glht(displayf.aov, linfct=mcp(panel="Tukey")),
s=multicomp(displayf.aov, "panel", plot=FALSE))
zapsmall(
if.R(r=t(tmp$linfct),
s=tmp$lmat)
)

## MMC plot
displayf.mmc <-
if.R(r=glht.mmc(displayf.aov, linfct=mcp(panel="Tukey"), focus="panel", lmat.rows=5:6),
s=multicomp.mmc(displayf.aov, "panel", lmat.rows=6:8, plot=FALSE))
plot(displayf.mmc)

## orthogonal contrasts
zapsmall(mca.lmat <- displayf.mmc$mca$lmat)
panel.lmat <- cbind("3-12"=mca.lmat[,1] + mca.lmat[,2],
"1-2"=mca.lmat[,3])
displayf.mmc <-
if.R(r=glht.mmc(displayf.aov, linfct=mcp(panel="Tukey"), focus="panel",
lmat.rows=5:6, lmat=panel.lmat),
s=multicomp.mmc(displayf.aov, "panel", lmat.rows=6:8,
lmat=panel.lmat, plot=FALSE))
plot(displayf.mmc)

## split plot design with tiebreaker plot

##
## This example is based on the query by Tomas Goicoa to R-news
## http://article.gmane.org/gmane.comp.lang.r.general/76275/match=goicoa
## It is a split plot similar to the one in HH Section 14.2 based on
## Yates 1937 example. I am using the Goicoa example here because its
## MMC plot requires a tiebreaker plot.

maiz <- read.table(hh("datasets/maiz.dat"), header=TRUE)

maiz$hibrido <- factor(maiz$hibrido,
levels=c("P3747","P3732","Mol17","A632","LH74"))
maiz$nitrogeno <- factor(maiz$nitrogeno)
position(maiz$nitrogeno) <- c(1, 2.5, 4, 5.5) ## forces class="ordered"

interaction2wt(yield ~ hibrido+nitrogeno+bloque, data=maiz)

interaction2wt(yield ~ hibrido+nitrogeno, data=maiz)

maiz.aov <- aov(yield ~ nitrogeno*hibrido + Error(bloque/nitrogeno), data=maiz)

summary(maiz.aov)
summary(maiz.aov,
split=list(hibrido=list(P3732=1, Mol17=2, A632=3, LH74=4)))

## multicomp(maiz.aov, focus="hibrido") ## can't use 'aovlist' objects

## glht(maiz.aov, linfct=mcp(hibrido="Tukey")) ## can't use 'aovlist' objects
mmc.mean 57

sapply(maiz[-1], contrasts)
if.R(r={
## R glht.mmc requires treatment contrasts
contrasts(maiz$nitrogeno) <- "contr.treatment"
sapply(maiz[-1], contrasts)
},
s={})

## Both R and S-Plus require aov, not aovlist

maiz2.aov <- aov(terms(yield ~ bloque*nitrogeno + hibrido/nitrogeno,
keep.order=TRUE), data=maiz)
summary(maiz2.aov)

if.R(s={
maiz2.mca <- multicomp(maiz2.aov, focus="hibrido")
## plot(maiz2.mca)
dimnames(maiz2.mca$lmat)[[1]]
maiz2.mmc <- multicomp.mmc(maiz2.aov, focus="hibrido",
lmat.rows=16:20, plot=FALSE)
old.mar <- par(mar=c(15,4,4,7)+.1)
plot(maiz2.mmc)
par(mar=c(2,4,28,7)+.1, new=TRUE)
old.cex <- par(cex=.8)
plot(maiz2.mmc$mca, col.signif=8, lty.signif=1, xlabel.print=FALSE,
xaxs="d", plt=par()$plt+c(0,0,-.25,.05), xrange.include=c(-30,40))
par(old.cex)
par(old.mar)
},r={
maiz2.mca <- glht(maiz2.aov, linfct=mcp(hibrido="Tukey"))
dimnames(maiz2.mca$linfct)[[2]]
maiz2.mmc <- glht.mmc(maiz2.aov, linfct=mcp(hibrido="Tukey"), lmat.rows=9:12)
old.oma <- par(oma=c(12,3,0,4))
plot(maiz2.mmc)
par(oma=c(0,3,22,4), new=TRUE)
plot(maiz2.mmc$mca,
xlim=par()$usr[1:2], xaxs="i",
main="", xlab="", cex.axis=.7)
par(old.oma)
})

mmc.mean MMC (mean–mean multiple comparisons) plots from the sufficient

statistics for a one-way design.

Description
Constructs a "mmc.multicomp" object from the sufficient statistics for a one-way design. The
object must be explicitly plotted.
58 mmc.mean

Usage

multicomp.mean(group, n, ybar, s, alpha=.05, ## S-Plus

ylabel="ylabel", focus.name="focus.factor", plot=FALSE,
lmat, labels=NULL, ...,
df=sum(n) - length(n),
sigmahat=(sum((n-1)*s^2) / df)^.5)

multicomp.mmc.mean(group, n, ybar, s, ylabel, focus.name, ## S-Plus

lmat,
...,
comparisons="mca",
lmat.rows=seq(length=length(ybar)),
ry,
plot=TRUE,
crit.point,
iso.name=TRUE,
estimate.sign=1,
x.offset=0,
order.contrasts=TRUE,
method="tukey",
df=sum(n)-length(n),
sigmahat=(sum((n-1)*s^2)/df)^.5)

Arguments
group character vector of levels
n numeric vector of sample sizes
ybar vector of group means
s vector of group standard deviations
alpha Significance levels of test
ylabel name of response variable
focus.name name of factor
plot logical. Should the "mmc.multicomp" object be automatically plotted? ig-
nored in R.
lmat lmat from multicomp in S-Plus or t(linfct) from glht in R.
labels labels argument for multicomp in S-Plus. Not used in R.
method method for critical point calculation. This corresponds to method in S-Plus
multicomp and to type in R glht
df scalar, residual degrees of freedom
sigmahat sqrt(MSE) from the ANOVA table
... other arguments
comparisons argument to S-Plus multicomp only.
mmc.mean 59

estimate.sign, order.contrasts, lmat.rows

See lmat.rows in mmc.
ry See argument ry.mmc in plot.mmc.multicomp.
crit.point See argument crit.point in S-Plus multicomp. The equivalent is not in
glht.
iso.name, x.offset
See plot.mmc.multicomp.

Value
multicomp.mmc.mean returns a "mmc.multicomp" object.
multicomp.mean returns a "multicomp" object.

Author(s)
Richard M. Heiberger <rmh@temple.edu>

## This is the S-Plus version

## See ?aov.sufficient for R

if.R(r={},
s={

pulmonary <- read.table(hh("datasets/pulmonary.dat"), header=TRUE,

60 mmc.mean

## multicomp object
pulmonary.mca <-
multicomp.mean(pulmonary$smoker,
pulmonary$n,
pulmonary$FVC,
pulmonary$s,
ylabel="pulmonary",
focus="smoker")

pulmonary.mca
## lexicographic ordering of contrasts, some positive and some negative
plot(pulmonary.mca)

pulm.lmat <- cbind("npnl-mh"=c( 1, 1, 1, 1,-2,-2), ## not.much vs lots

## mmc.multicomp object
pulmonary.mmc <-
multicomp.mmc.mean(pulmonary$smoker,
pulmonary$n,
pulmonary$FVC,
pulmonary$s,
ylabel="pulmonary",
focus="smoker",
lmat=pulm.lmat,
plot=FALSE)

old.par <- par(mar=c(5,4,4,4)+.1)

## pairwise comparisons
plot(pulmonary.mmc, print.mca=TRUE, print.lmat=FALSE)

## tiebreaker plot, with contrasts ordered to match MMC plot,

## with all contrasts forced positive and with names also reversed,
## and with matched x-scale.
plot(pulmonary.mmc$mca, col.signif='red', lty.signif=1, xlabel.print=FALSE,
xaxs="d", plt=par()$plt+c(0,0,-.25,.05), xrange.include=c(-1, 1))
multicomp.order 61

## orthogonal contrasts
plot(pulmonary.mmc, print.lmat=TRUE, col.lmat.signif='blue', col.iso='gray')

## pairwise and orthogonal contrasts on the same plot

plot(pulmonary.mmc, print.mca=TRUE, print.lmat=TRUE,
col.mca.signif='red', col.lmat.signif='blue', col.iso='gray',
lty.lmat.not.signif=2)

par(old.par)
})

multicomp.order Update a multicomp object by ordering its contrasts.

Description
Update a multicomp object by ordering its contrasts. The default sort.by = "height" matches
the order in the MMC plot. An alternate sort.by = "estimate" matches the order of the
half-normal plot. Or the argument sort.order can be used to specify any other order.

Usage
multicomp.order(mca, sort.by = "height", sort.order = NULL)

multicomp.label.change(x, old="adj", new="new", how.many=2)

## S3 method for class 'multicomp':

multicomp.label.change(x, old="adj", new="new", how.many=2)

## S3 method for class 'mmc.multicomp':

multicomp.label.change(x, old="adj", new="new", how.many=2)

Arguments
mca "multicomp" object. This is the result of multicomp in S-Plus or the result
from applying as.multicomp to a "glht" object in R.
sort.by Either "height" or "estimate".
sort.order Vector of indices by which the contrasts are to be sorted. When sort.order
in non-NULL, it is used.
x "multicomp" object.
old character string to be removed from contrast names.
new replacement character string to be inserted in contrast names.
how.many number of times to make the replacement.
62 multicomp.order

Value
The result is a "multicomp" object containing the same contrasts as the argument. multicomp.order
sorts the contrasts (and renames them consistently) according to the specifications. multicomp.label.change
changes the contrast names according to the specifications.
When sort.by=="height", sort the contrasts by the reverse order of the heights. This provides
a "multicomp" object that will be plotted by plot.multicomp in the same order used by
plot.mmc.multicomp. If there is not "height" component, the original "multicomp"
object is returned.
When sort.by=="estimate", sort the contrasts by the reverse order of the contrast estimates.
This provides the same order as the half-normal plot.
When sort.order in non-NULL, sort the contrasts in that order.

Note
S-Plus use the multicomp functions and R uses the multcomp package.

Author(s)
Richard M. Heiberger <rmh@temple.edu>

See Also
MMC, as.glht in R, multicomp.reverse

Examples
## continue with the example in glht.mmc in R, or multicomp.mmc in S-Plus
catalystm <- read.table(hh("datasets/catalystm.dat"), header=FALSE,
col.names=c("catalyst","concent"))
catalystm$catalyst <- factor(catalystm$catalyst, labels=c("A","B","C","D"))
catalystm1.aov <- aov(concent ~ catalyst, data=catalystm)

if.R(r={
catalystm.mca <-
glht(catalystm1.aov, linfct = mcp(catalyst = "Tukey"))
print(confint(catalystm.mca))

catalystm.mmc <-
glht.mmc(catalystm1.aov, linfct = mcp(catalyst = "Tukey"))
## the contrasts have been ordered by height (see ?MMC),
## which in this example corresponds to sort.order=c(1,2,4,3,5,6),
multicomp.reverse 63

## and reversed, to make the contrast Estimates positive.

print(as.glht(catalystm.mmc$mca))

## ## For consistency with the S-Plus example,

## ## we change all factor level "A" to "control".
## as.glht(multicomp.label.change(catalystm.mmc$mca, "A", "control"))
},s={
catalystm.mca <-
multicomp(catalystm1.aov, method="Tukey")
print(catalystm.mca)

catalystm.mmc <-
multicomp.mmc(catalystm1.aov, method="Tukey", plot=FALSE)
## the contrasts have been ordered by height (see ?MMC),
## which in this example corresponds to sort.order=c(1,2,4,3,5,6),
## and reversed, to make the contrast Estimates positive.
print(catalystm.mmc$mca)

## S-Plus multicomp already uses simple names. This function is

## therefore used in more complex two-way ANOVA examples. We illustrate
## here by changing all factor level "A" to "control".
print(multicomp.label.change(catalystm.mmc$mca, "A", "control"))

})

multicomp.reverse Force all comparisons in a "multicomp" object to have the same sign.

Description
Force all comparisons in a "multicomp" object to have the same sign. If the contrast "A-B" has
a negative estimate, reverse it show the contrast "B-A" with a positive estimate.

Usage
multicomp.reverse(y, estimate.sign = 1, ...)

Arguments
y "multicomp" object
estimate.sign
If estimate.sign==1, reverse the negatives. If estimate.sign==-1,
reverse the positives. Both the names of the comparisons and the numerical
values are reversed. If estimate.sign==0, return the argument.
... other arguments not used.

Value
The result is a "multicomp" object containing the same contrasts as the argument but with the
sign of the contrasts changed as needed.
64 norm.curve

Note
S-Plus use the multicomp functions and R uses the multcomp package.

Author(s)
Richard M. Heiberger <rmh@temple.edu>

norm.curve plot a normal or a t-curve with both x and z axes.

Description
Plot a normal curve or a t-curve with both x (with mean and se as specified) and z or t (mean=0,
se=1) axes. Shade a region for rejection region, acceptance region, confidence interval. The density
axis is marked in units appropriate for the z or t axis. The existence of any of the arguments
se, sd, n forces dual x and (z or t) scales. When none of these arguments are used, the main
title defaults to "Standard Normal Density N(0,1)" and only the z scale is printed. A
second density curve, appropriate for an alternative hypothesis is displayed when the argument
axis.name="z1" is specified. The shaded area is printed on the plot.
When the optional argument df.t is specified, then a t-distribution with df.t degrees of freedom
is plotted.
norm.observed plots a vertical line with arrowhead markers at the location of the observed xbar.

Usage
norm.setup(xlim.in=c(-2.5,2.5),
ylim.in = c(0, 0.4)/se,
mean=0,
main.in=main.calc,
se=sd/sqrt(n), sd=1, n=1,
norm.curve 65

df.t=NULL,
...)

norm.curve(mean=0, se=sd/sqrt(n),
critical.values=mean + se*c(-1, 1)*z.975,
z=do.call("seq",
as.list(c((par()$usr[1:2]-mean)/se, length=109))),
shade, col=par("col"),
axis.name=ifelse(is.null(df.t) || df.t == Inf, "z", "t"),
second.axis.label.line=3,
sd=1, n=1,
df.t=NULL,
...)

norm.observed(xbar, t.xbar, col="blue")

Arguments
xlim.in, ylim.in
xlim, ylim. Defaults to correct values for standard Normal(0,1). User must
set values for other mean and standard error.
mean Mean of the normal distribution in xbar-scale, used in calls to dnorm.
se standard error of the normal distribution in xbar-scale, used in calls to dnorm.
sd, n standard deviation and sample size of the normal distribution in x-scale. These
may be used as an alternate way of specifying the standard error se.
df.t Degrees of freedom for the t distribution. When df.t is NULL, the normal
distribution is used.
critical.values
Critical values in xbar-scale. A scalar value implies a one-sided test. A vector
of two values implies a two-sided test.
main.in Main title. Default value is: if (is.null(df.t)) ## normal ifelse(!(missing(se)
&& missing(sd) && missing(n)), paste("normal density:
se =", round(se,3)), "Standard Normal Density N(0,1)")
else { ## t distribution if (length(df.t) != 1) stop("df.t
must have length 1") ifelse(!(missing(se) && missing(sd)
&& missing(n)), paste("t density: se = ", round(se,3),
", df = ", df.t, sep=""), paste("t density, df =", df.t))
}
z z-values (standardized to N(0,1)) used as base of plot.
shade Valid values for shade are "right", "left", "inside", "outside", "none". Default is
"right" for one-sided critical.values and "outside" for two-sided critical values.
col color of the shaded region and the area of the shaded region.
axis.name defaults to "z" for the standard normal scale centered on the null hypothesis
value of the mean or to "t" for the t distribution with df.t degrees of freedom.
For alternative hypotheses, the user must specify either "z1" or "t1" for the
standard normal scale, or t distibution with df.t degrees of freedom, centered
on the alternate hypothesis value of the mean.
66 norm.curve

second.axis.label.line
Defaults to 3. Normally not needed. When two curves are drawn, one normal
and one t, then the second curve needs a different label for the y-axis. Set this
value to 4 to avoid overprinting.
xbar xbar-value of the observed data.
t.xbar t-value of the observed data under the null hypothesis.
... Other arguments which are ignored.

Author(s)
Richard M. Heiberger <rmh@temple.edu>

Examples
old.par <- par(oma=c(4,0,2,5), mar=c(7,7,4,2)+.1)

norm.setup()
norm.curve()

norm.setup(xlim=c(75,125), mean=100, se=5)

norm.curve(100, 5, 100+5*(1.645))
norm.observed(112, (112-100)/5)

norm.setup(xlim=c(75,125), mean=100, se=5)

norm.curve(100, 5, 100+5*(-1.645), shade="left")

norm.setup(xlim=c(75,125), mean=100, se=5)

norm.curve(mean=100, se=5, col='red')

norm.setup(xlim=c(75,125), mean=100, se=5)

norm.curve(100, 5, 100+5*c(-1.96, 1.96))

norm.setup(xlim=c(-3, 6))
norm.curve(crit=1.645, mean=1.645+1.281552, col='green',
shade="left", axis.name="z1")
norm.curve(crit=1.645, col='red')

norm.setup(xlim=c(-6, 12), se=2)

norm.curve(crit=2*1.645, se=2, mean=2*(1.645+1.281552),
col='green', shade="left", axis.name="z1")
norm.curve(crit=2*1.645, se=2, mean=0,
col='red', shade="right")

par(mfrow=c(2,1))
norm.setup()
norm.curve()
mtext("norm.setup(); norm.curve()", side=1, line=5)
norm.setup(n=1)
norm.curve(n=1)
mtext("norm.setup(n=1); norm.curve(n=1)", side=1, line=5)
par(mfrow=c(1,1))
norm.curve 67

par(mfrow=c(2,2))

## naively scaled,
## areas under the curve are numerically the same but visually different
norm.setup(n=1)
norm.curve(n=1)
norm.observed(1.2, 1.2/(1/sqrt(1)))
norm.setup(n=2)
norm.curve(n=2)
norm.observed(1.2, 1.2/(1/sqrt(2)))
norm.setup(n=4)
norm.curve(n=4)
norm.observed(1.2, 1.2/(1/sqrt(4)))
norm.setup(n=10)
norm.curve(n=10)
norm.observed(1.2, 1.2/(1/sqrt(10)))
mtext("areas under the curve are numerically the same but visually different",
side=3, outer=TRUE)

## scaled so all areas under the curve are numerically and visually the same
norm.setup(n=1, ylim=c(0,1.3))
norm.curve(n=1)
norm.observed(1.2, 1.2/(1/sqrt(1)))
norm.setup(n=2, ylim=c(0,1.3))
norm.curve(n=2)
norm.observed(1.2, 1.2/(1/sqrt(2)))
norm.setup(n=4, ylim=c(0,1.3))
norm.curve(n=4)
norm.observed(1.2, 1.2/(1/sqrt(4)))
norm.setup(n=10, ylim=c(0,1.3))
norm.curve(n=10)
norm.observed(1.2, 1.2/(1/sqrt(10)))
mtext("all areas under the curve are numerically and visually the same",
side=3, outer=TRUE)

par(mfrow=c(1,1))

## t distribution
mu.H0 <- 16
se.val <- .4
df.val <- 10
crit.val <- mu.H0 - qt(.95, df.val) * se.val
mu.alt <- 15
obs.mean <- 14.8

alt.t <- (mu.alt - crit.val) / se.val

norm.setup(xlim=c(12, 19), se=se.val, df.t=df.val)
norm.curve(crit=crit.val, se=se.val, df.t=df.val, mean=mu.alt,
col='green', shade="left", axis.name="t1")
norm.curve(crit=crit.val, se=se.val, df.t=df.val, mean=mu.H0,
col='gray', shade="right")
norm.observed(obs.mean, (obs.mean-mu.H0)/se.val)
68 npar.arma

## normal
norm.setup(xlim=c(12, 19), se=se.val)
norm.curve(crit=crit.val, se=se.val, mean=mu.alt,
col='green', shade="left", axis.name="z1")
norm.curve(crit=crit.val, se=se.val, mean=mu.H0,
col='gray', shade="right")
norm.observed(obs.mean, (obs.mean-mu.H0)/se.val)

## normal and t
norm.setup(xlim=c(12, 19), se=se.val, main="t(6) and normal")
norm.curve(crit=15.5, se=se.val, mean=16.3,
col='gray', shade="right")
norm.curve(crit=15.5, se.val, df.t=6, mean=14.7,
col='green', shade="left", axis.name="t1", second.axis.label.line=4)
norm.curve(crit=15.5, se=se.val, mean=16.3,
col='gray', shade="none")

norm.setup(xlim=c(12, 19), se=se.val, main="t(6) and normal")

norm.curve(crit=15.5, se=se.val, mean=15.5,
col='gray', shade="right")
norm.curve(crit=15.5, se=se.val, df.t=6, mean=15.5,
col='green', shade="left", axis.name="t1", second.axis.label.line=4)
norm.curve(crit=15.5, se=se.val, mean=15.5,
col='gray', shade="none")

par(old.par)

npar.arma Count the number of parameters in an ARIMA model specification.

Description
Count the number of parameters in an ARIMA model specification. When arima==FALSE, just
the AR and MA parameters are counted. When arima==TRUE, then the number of difference
parameters are also included.

Usage
npar.arma(x, arima=FALSE)
npar.sarma(model, arima=FALSE)
npar.rarma(arma, arima=FALSE)

Arguments
x An "arima" object in S-Plus or a "Arima" object in R.
model A model specification in the S-Plus style.
objip 69

arma A arma specification in the R style

arima Logical. TRUE is number of differencings is to be counted.

Value
A scalar number giving the count.

Author(s)
Richard M. Heiberger (rmh@temple.edu)

Examples
co2.arima <-
if.R(s=
arima.mle(co2, list(list(order=c(0,1,1)),
list(order=c(0,1,1), period=12)))
,r=
arima(co2,
order=c(0,1,1),
seasonal=list(order=c(0,1,1), period=12))
)

npar.arma(co2.arima)

npar.arma(co2.arima, arima=TRUE)

npar.sarma(list(list(order=c(0,1,1)),
list(order=c(0,1,1), period=12)))

npar.sarma(list(list(order=c(0,1,1)),
list(order=c(0,1,1), period=12)),
arima=TRUE)

if.R(s={},
r=npar.rarma(co2.arima$arma)
)
if.R(s={},
r=npar.rarma(co2.arima$arma,
arima=TRUE)
)

objip loop through all attached directories looking for pattern

Description
Loop objects() through all attached directories (items in the search() list) looking for a
regular expression pattern.
70 odds.ratio

Usage
objip(pattern, where = search(), frame=NULL)

Arguments
pattern Character string containing a regular expression that is used to list only a subset
of the objects. Only names matching ’pattern’ are returned.
where an object defining a database in the search list.
frame In S-Plus, an integer giving the frame number. In R, frame is ignored.

Value
A list of 0 or more character vectors. Each character vector has the name of one of the items
in the search() list. Each character vector contains the names of the objects in the specified
environment which match the pattern. If there are no matching names in an environment, then
the corresponding character vector is removed from the result.

Author(s)
Richard M. Heiberger <rmh@temple.edu>

See Also
objects in R, objects in S-Plus.

Examples
objip("qq")
objip("^qq")
objip("qq$")

odds.ratio Calculate or plot the odds ratio for a 2x2 table of counts.

Description
Calculate or plot the odds ratio for a 2x2 table of counts. The plot shows the confidence intervals
on the probability of row2 for fixed odds ratio and specified probability for row1.

Usage
odds.ratio(x, alpha = 0.05)

plot.odds.ratio(x,
ylab = "prob(col1 | row1)", xlab = "prob(col1 | row2)",
alpha = c(0.05, 0.5),
legend.x=1.05,
oma=c(0,0,0,5), ...)
odds.ratio 71

Arguments

x 2 x 2 table of counts
alpha Significance levels of test. odds.ratio requires a single number in the range
[0,1]. plot.odds.ratio accepts more than one number on the range [0,1]
and draws confidence lines at each value.
xlab, ylab x- and y-labels for the plot. Sensible defaults are generated.
... other arguments, currently ignored.
legend.x x position of left-hand side of legend.
oma outer margin par()$oma, needed to make room for legend.

Value

plot.odds.ratio draws a plot and invisibly returns the same list as odds.ratio for the first
value of alpha. odds.ratio returns the list:

p1, p2 proportion of each row total observed in the first column.

omega1, omega2
odds for each row, p/(1-p)
psihat odds ratio, omega2/omega1
s.ln.psihat standard deviation of ln(psihat)
ci.ln.psihat confidence interval for ln(psihat) using normal approximation
ci.psihat confidence interval for psihat calculated ase p(ci.ln.psihat)
prob1 seq(0,1,.05), set of p1 values for plotting.
odds1 p1/(1-p1)
odds2 odds for the second row needed to retain psihat with the specified odds1,
calculated as odds1*psihat.
ci.odds2 confidence interval for odds2
prob2 odds2 / (odds2+1)
ci.prob2 ci.odds2 / (ci.odds2+1)

Author(s)

Richard M. Heiberger <rmh@temple.edu>

References

Examples
glasses <- read.table(hh("datasets/glasses.dat"), header=FALSE)

glasses <- matrix(unlist(glasses),

nrow=2,
dimnames=list(
c("glasses","no.glasses"),
c("delinq","non.del")))

## draw the iso-odds ratio plot with 50% CI and 95% CI,
## invisibly return the 95% CI
plot.odds.ratio(glasses)

orthog.complete Construct an orthogonal matrix which is an arbitrary completion of

the column space of the input set of columns.

Description

Usage
orthog.complete(x, normalize=TRUE, abs2.rows=1:nrow(x),
Int=TRUE, drop.Int=Int)

orthog.construct(y, x, x.rows, normalize=FALSE)

Arguments
x For orthog.complete, an n-row matrix of one or more columns. For orthog.construct,
a set of contrasts for a factor.
y matrix of coefficients specifying the linear combinations estimated. This will
usually be the lmat from an S-Plus "multicomp" object or the linfct
component from an R "glht" object.
normalize, abs2.rows, x.rows
The default normalizes the sum of squares of the rows in abs2.rows or x.rows
to 1. The optional value normalize="abs2" scales the rows in abs2.rows
or x.rows to make the sum of all positive value equal 1 and the sum of all neg-
ative values equal -1. Together, the sum of the absolute values ofthe specified
rows in each column is 2.
Int logical. Default TRUE means make all columns orthogonal to the constant col-
umn (Intercept in regression terminology). The alternative is to use only the
columns in the input matrix x.
drop.Int logical. The default is to drop the constant column and to keep all columns when
the constant is not automatically generated.
orthog.complete 73

Details
This function is based on qr.Q. The input matrix x has n rows and an arbitrary non-zero number
of columns. The result is an n by n orthogonal matrix. By default the first column of the result
is constant and is not returned. The second column of the result is orthogonal to the first result
column. Together the first two result columns span the space of the constant column and the first
input column. The third result column is orthogonal to the first two result columns and the the
three result columns together span the space of the constant column and the first two inout columns.
Similarly for the remaining result columns. Result columns beyond the number of input columns
are constructed as an arbitrary orthogonal completion.
If the input columns are orthogonal to each other and to the constant column, then the result columns
are rescaled versions of the input columns.
Optionally (drop.Int=FALSE), the constant column can be returned.
By default the columns are scaled to have sum of squares equal 1. If normalize="abs2", they
are scaled to make the sum of all positive value equal 1 and the sum of all negative values equal -1.
Together, the sum of the absolute values of each column is 2.
If the input is a set of columns from a contrast matrix for a design that has multiple terms, the
abs2.rows argument is used to specify which rows are to be included in the normalization. These
will normally be rows associated with one of the main effects.

Value
Matrix of orthogonal columns.

Author(s)
Richard M. Heiberger <rmh@temple.edu>

## used in MMC plots

tmp <- data.frame(y=rnorm(12),
a=factor(c("u","u","u","u",
"v","v","v","v",
"w","w","w","w")))
tmp.aov <- aov(y ~ a, data=tmp)
lmat <- if.R(
s=multicomp(tmp.aov, focus="a")$lmat,
r={lmat.reduced <- t(glht(tmp.aov, linfct=mcp(a="Tukey"))$linfct)
rbind(lmat.reduced, AU=-apply(lmat.reduced[-1,], 2, sum))
})
zapsmall(lmat)

lmat.complete <- orthog.complete(lmat, abs2.rows=-1,

normalize="abs2",
drop.Int=FALSE)[,1:3]
zapsmall(lmat.complete)
if.R(r=zapsmall(lmat.complete[-4,]),
s={})

panel.acf Panel functions for tsdiagplot.

Description
Panel functions for tsdiagplot.

Usage
panel.acf(..., n.used)
panel.std.resid(...)
panel.gof(...)

Arguments
... standard arguments to panel functions.
n.used number of lags. THis number is needed to calculate the confidence interval
which is 2//sqrt(n.used).

Author(s)
Richard M. Heiberger (rmh@temple.edu)
panel.bwplot.intermediate.hh 75

panel.bwplott Extension to S-Plus trellis to allow transposed plots.

Description
Extension to S-Plus trellis to allow transposed plots. All x - and y-components of the trellis ob-
ject are interchanged. This function is not needed in R as lattice has a horizontal argument in its
definitions.

Usage
panel.bwplott(x, y, box.ratio = 1,
font = box.dot$font, pch = box.dot$pch, cex = box.dot$cex,
col = box.dot$col, ..., transpose=FALSE)

Arguments
x, y, box.ratio, font, pch, cex, col, ...
See panel.bwplot in S-Plus or panel.bwplot in R.
transpose logical. If FALSE, the plot is printed in the default orientation. If TRUE, the
x- and y-components of the trellis object are interchanged. This has the effect,
for example, of displaying vertical boxplots instead of the default horizontal
boxplots.

Value
The function is used for its side effect of drawing boxplots in a trellis panel.

Note
This function is not needed in R. If it is used and

Author(s)
Richard M. Heiberger <rmh@temple.edu>

See Also
bwplot in S-Plus or bwplot in R.

Examples
panel.cartesian 77

panel.cartesian trellis panel function, with labeled rows and columns and without strip
labels.

Description

trellis panel function, with labeled rows and columns and without strip labels. Designed for use
with the ladder of powers plot.

Usage

panel.cartesian(x, y,
x.label=unique(panel.labels[,"x"]),
y.label=unique(panel.labels[,"y"]),
group.label.side="",
axis3.line=1,
xg.label, yg.label, g.cex=.7,
rescale=list(x=TRUE,y=TRUE), ...,
browser.on=FALSE)

Arguments

x, y x and y as for any other panel function

x.label labels for the columns of the scatterplot matrix
y.label labels for the rows of the scatterplot matrix
axis3.line The x.label doesn’t always show up in the right place. This allows the user
to adjust it’s position.
group.label.side
c("","left","top"), when the plotting formula is conditioned on a group
factor, the levels of the group are displayed in the margins of the plot. The
appearance depends on the setting of the trellis between argument. Getting it
to look good for any given plot requires experimentation. Since it is redundant
with the information in the strip labels, leaving it at the default "" is often the
best thing to do..
xg.label group labels for rows of the scatterplot matrix
yg.label group labels for rows of the scatterplot matrix
g.cex cex for the group labels
rescale alternate way to get something similar to relation="free"
... other arguments
browser.on logical, normally FALSE. This is a debugging tool. When TRUE, the browser()
is turned on at various critical points.
78 panel.cartesian

rent <- read.table(hh("datasets/rent.dat"), ## Weisberg's file alr162

col.names=c("rnt.alf","rnt.till",
"cow.dens","prop.past","lime"))
rent$lime <- factor(rent$lime, labels=c("no.lime","lime"))
rent$alf.till <- rent$rnt.alf / rent$rnt.till

rent.lm <- lm(rnt.alf ~ rnt.till + cow.dens + lime, data=rent)

rent$resid.rent <- resid(rent.lm)

xysplom(resid.rent ~ rnt.till + cow.dens | lime, data=rent,

layout=c(2,2))

xysplom(resid.rent ~ rnt.till + cow.dens | lime, data=rent,

layout=c(2,2),
xlab="", ylab="",
x.label="", y.label="",
group.label.side="",
par.strip.text=list(cex=1.2),
panel=panel.cartesian,
axis3.line=2.4,
scales=list(
relation="same",
alternating=FALSE, labels=FALSE, ticks=FALSE),
between=list(x=1, y=3))

xysplom(resid.rent ~ rnt.till + cow.dens | lime, data=rent,

layout=c(2,2),
xlab="", ylab="",
x.label="", y.label="",
group.label.side="",
par.strip.text=list(cex=1.2),
panel=panel.cartesian,
axis3.line=3.6,
scales=list(
relation="same",
alternating=FALSE, labels=FALSE, ticks=FALSE),
rescale=list(x=FALSE, y=FALSE),
between=list(x=1, y=3))

xysplom(resid.rent ~ rnt.till + cow.dens | lime, data=rent,

panel.ci.plot 79

layout=c(2,2),
xlab="", ylab="",
x.label="", y.label="",
group.label.side="",
par.strip.text=list(cex=1.2),
panel=panel.cartesian,
axis3.line=3.6,
scales=list(
relation="free",
alternating=FALSE, labels=FALSE, ticks=FALSE),
between=list(x=1, y=3))

xysplom(resid.rent ~ rnt.till + cow.dens | lime, data=rent,

layout=c(2,2),
xlab="", ylab="",
y.label="resid",
group.label.side="top",
par.strip.text=list(cex=1.2),
panel=panel.cartesian,
axis3.line=3.6,
scales=list(alternating=FALSE, labels=FALSE, ticks=FALSE),
rescale=list(x=FALSE, y=FALSE),
between=list(x=1, y=5))

panel.ci.plot Default Panel Function for ci.plot

Description
This is the default panel function for ci.plot.

Usage
panel.ci.plot(x, y, newdata, newfit = newfit, ...)

Arguments
x Observed values of predictor variable.
y Observed values of response variable.
newdata x values for which predictions are calculated.
newfit data.frame containing six components: "fit", "se.fit", "residual.scale",
"df", "ci.fit", "pi.fit". In S-Plus these are the output from the predict.lm
function. In R they are a rearrangement of the output of the predict.lm func-
tion.
... other arguments passed to panel.xyplot.

Author(s)
Richard M. Heiberger <rmh@temple.edu>
80 panel.dotplot.tb

panel.dotplot.tb Dotplot with evenly spaced tiebreakers.

Description
Dotplot with evenly spaced tiebreakers. Multiple hits on a specific x value are stacked.

Usage
panel.dotplot.tb(x, y, factor=.1,
jitter.data=TRUE, horizontal=TRUE,
max.freq=max(sapply(subsets, length)),
...)

Arguments
x, y See dotplot.
factor jitter factor, see stripplot. Increment is factor/max.freq where max.freq
is the maximum number of duplicates of any x value in any y group.
jitter.data, horizontal
Always TRUE.
max.freq maximum number of observation at any combination of response values, fac-
tor levels, and group levels. If the formula includes one or more conditioning
factors, then the user is responsible for providing a value for max.freq.
... Other arguments for dotplot.

Details
Creates (possibly grouped) Dotplot of x against y. y is the ‘factor’.

Warning
If the formula includes one or more conditioning factors, then the user is responsible for providing a
value for max.freq. The default behavior is a different max.freq for each panel in a multi-panel
display.

Author(s)
Richard M. Heiberger
Maintainer: Richard M. Heiberger <rmh@temple.edu>
panel.interaction2wt 81

Examples
x <- c(1,1,2,2,2,5,4,2,1,5)
y <- factor(letters[rep(1:2, 5)])

dotplot(x, panel=panel.dotplot.tb)
dotplot(x, panel=panel.dotplot.tb, factor=.2)
dotplot(y ~ x, panel=panel.dotplot.tb)
dotplot(y ~ x, panel=panel.dotplot.tb, cex=1.5, factor=.15)

quiz <- data.frame(scores=sample(10, 360, replace=TRUE),

date=rep(rep(c("0902", "0916", "0930"), c(40,40,40)), 3),
section=rep(
c("Stat1-3", "Stat1-5", "Stat1-8"),
c(120,120,120)))

dotplot(date ~ scores | section, data=quiz,

panel=panel.dotplot.tb, factor=.5)

dotplot(date ~ scores | section, data=quiz,

panel=panel.dotplot.tb, factor=.5,
layout=c(1,3), between=list(y=1),
main='Three quizzes for three sections of Stat 1')

## If the formula includes one or more conditioning factors, then the

## user is responsible for providing a value for the argument max.freq
##
a <- rep(1, 10)
z <- c(1,1,2,2,2,3,2,3,1,1)
g <- LETTERS[c(1,1,1,1,1,2,2,2,2,2)]

print(split=c(1,1,2,1), more=TRUE,
dotplot( a ~ z | g, panel=panel.dotplot.tb,
factor=.6, cex=1.5, layout=c(2,1),
main="different scaling in each panel")
)

print(split=c(2,1,2,1), more=FALSE,
dotplot( a ~ z | g, panel=panel.dotplot.tb, max.freq=3,
factor=.6, cex=1.5, layout=c(2,1),
main="same scaling in each panel")
)

panel.interaction2wt
Plot all main effects and twoway interactions in a multifactor design
82 panel.interaction2wt

Description
This is the panel function for interaction2wt. The main diagonal displays boxplots for the
main effects of each factor. The off-diagonals show the interaction plots for each pair of factors. The
i,j panel shows the same factors as the j,i but with the trace- and x-factor roles interchanged.

Usage
panel.interaction2wt(x, y, subscripts,
responselab, trace.values,
factor.levels, factor.position,
fun = mean,
se,
...,
box.ratio,
simple=FALSE,
simple.offset,
simple.scale,
data.x,
col.by.row=TRUE,
key.in=NULL)

strip.interaction2wt(which.given, which.panel, var.name,

factor.levels, shingle.intervals,
strip.names = c(TRUE, TRUE), style = 1, ...)

Arguments
panel.interaction2wt arguments:

x levels of x-factor
y Summary value of response variable at each level of x- and trace-factors.
subscripts used to get the right set of response values for the summary statistics on the
off-diagonals
responselab Character name of response variable, defaults to the name of the response vari-
able.
trace.values levels of trace-factor
fun Summary function, defaults to mean
se standard errors to be passed to panel.intxplot. se Missing, logical, or a
numeric vector. If missing or FALSE, standard errors are not plotted. If TRUE,
the standard errors are calculated from the sufficient statistics for each group as
the groups standard deviation divided by the square root of the group observation
count. If a numeric vector, it is evaluated in the environment of the sufficient
statistics.
box.ratio passed to panel.bwplot.intermediate.hh
... extra arguments, primarily color, to be passed to panel.bwplot.intermediate.hh
panel.pairs.hh 83

key.in arguments to be passed through to the key for the trace-factor in each row of
the display.
factor.position
"position" attribute of factor.
simple logical. TRUE if simple effects are to be displayed.
simple.offset, simple.scale
named list of offset and scale for the response and trace factors. See interaction.positioned
for their use.
data.x data.frame containing factors from the input data.frame
col.by.row logical. If TRUE (the default), simple effects plots color the simple effects on
the main diagonals in the same color as the trace levels in their row. If FALSE,
then simple effects are colored to match the x levels in their column.
which.given, which.panel, var.name, factor.levels, shingle.intervals
see documentation for strip.default
strip.names Force strip.names=TRUE
style force style=1

Author(s)
Richard M. Heiberger <rmh@temple.edu>

See Also
interaction2wt, panel.bwplot.intermediate.hh

panel.pairs.hh Function based on S-Plus panel.pairs to add the subpanel.scales and

panel.cex arguments.

Description
Function based on S-Plus panel.pairs to add the subpanel.scales and panel.cex ar-
guments. In R, this is an alias for panel.pairs.

Usage
panel.pairs.hh(x, y, z, subscripts, pscales, subpanel = panel.splom,
varnames = dimnames(x)[[2]], ...,
subpanel.scales, panel.cex=par()$cex)
84 panel.xysplom

Arguments
x, y, z, subscripts, pscales, subpanel, varnames, ...
See splom in S-Plus.
subpanel.scales
Controls the size of the tick labels in the diagonal panel.
panel.cex Controls the size of the variable names in the diagonal panel.

Value
"trellis" object.

Author(s)
Richard M. Heiberger <rmh@temple.edu>

panel.xysplom panel method for xysplom.

Description
panel method for xysplom. It has a corr argument that is removed before sending the information
on to panel.xyplot.
partial.corr 85

Usage

panel.xysplom(corr, ...)

Arguments

corr logical. If TRUE, display the correlation and/or the regression coefficient for
lm(y ~ x) for each panel in an additional strip label.
... Remaining arguments to panel.xyplot.

Author(s)

Richard M. Heiberger <rmh@temple.edu>

See Also

xysplom

partial.corr partial correlations

Description

The partial correlation of x and y conditioning on z is the ordinary correlation of the residuals from
the regression of x on z and the regression of y on z.

Usage

partial.corr(vars, cond)

Arguments

vars matrix of data.frame of all the variables to be correlated.

cond matrix of data.frame of all the variables to be conditioned on.

Value

matrix of partial correlations of the numeric variables in the argument vars conditioning on the
numeric variables in cond.

Author(s)

Richard M. Heiberger <rmh@temple.edu>

86 plot.hov

Examples
if.R(r=
partial.corr(longley[,1:3], longley[,4:6])
,s=
partial.corr(longley.x[,1:3], longley.x[,4:6])
)

plot.hov Homogeneity of Variance Plot

Description
Oneway analysis of variance makes the assumption that the variances of the groups are equal.
Brown and Forsyth, 1974 present the recommended test of this assumption. The Brown and Forsyth
test statistic is the F statistic resulting from an ordinary one-way analysis of variance on the absolute
deviations from the median. The plot.hov function graphs the components of the Brown and
Forsyth test statistic.

Usage
plot.hov(x, data = sys.parent(), method = "bf", ## x is a formula
transpose.in = TRUE, ...)

## users will normally use the formula above and will not call the
## method directly.
plot.hov.bf(x, group, ## x is the response variable
y.name = deparse(substitute(x)),
group.name = deparse(substitute(group)),
transpose.in = TRUE, ...)

## users will normally use the formula above and will not call the
## panel function directly.
panel.hov(..., transpose = TRUE)

Arguments
x Formula appropriate for oneway anova in plot.hov. Response variable in
plot.hov.bf.
data data.frame
method Character string defining method. At this time the only recognized method is
"bf" for the Brown-Forsyth method.
transpose.in, transpose
Always TRUE in R. Normally TRUE in S-Plus to force vertical boxplots.
group factor.
y.name name of response variable, defaults to variable name in formula.
group.name name of factor, defaults to variable name in formula.
... additional arguments.
plot.mmc.multicomp 87

Value
"trellis" object with three panels containing boxplots for each group: The observed data "y",
the data with the median subtracted "y-med(y)", and the absolute deviations from the median
"abs(y-med(y))" The Brown and Forsyth test statistic is the F statistic resulting from an ordi-
nary one-way analysis of variance on the data points in the third panel.

Author(s)
Richard M. Heiberger <rmh@temple.edu>

References
Heiberger, Richard M. and Holland, Burt (2004b). Statistical Analysis and Data Display: An Inter-
mediate Course with Examples in S-Plus, R, and SAS. Springer Texts in Statistics. Springer. ISBN
0-387-40270-5.
Brown, M. B. and Forsyth, A. B. (1974). Robust tests for equality of variances. Journal of the
American Statistical Association, 69:364–367.

hov(wt.gain ~ diet, data=turkey)

plot.hov(wt.gain ~ diet, data=turkey)

plot.mmc.multicomp MMC (Mean–mean Multiple Comparisons) plot.

Description
MMC (Mean–mean Multiple Comparisons) plot.

Usage
## S3 method for class 'mmc.multicomp':
plot(x,
xlab="contrast value",
ylab=none$ylabel,
focus=none$focus,
main= main.method.phrase,
88 plot.mmc.multicomp

main2=main2.method.phrase,
main.method.phrase=
paste("multiple comparisons of means of", ylab),
main2.method.phrase=paste("simultaneous ",
100*(1-none$alpha),"% confidence limits, ",
method, " method", sep="" ),
ry.mmc=TRUE,
key.x=par()$usr[1]+ diff(par()$usr[1:2])/20,
key.y=par()$usr[3]+ diff(par()$usr[3:4])/3,
method=if (is.null(mca)) lmat$method else mca$method,
print.lmat=(!is.null(lmat)),
print.mca=(!is.null(mca) && (!print.lmat)),
iso.name=TRUE,
x.offset=0,
col.mca.signif="red", col.mca.not.signif="black",
lty.mca.signif=1, lty.mca.not.signif=6,
lwd.mca.signif=1, lwd.mca.not.signif=1,
col.lmat.signif="blue", col.lmat.not.signif="black",
lty.lmat.signif=1, lty.lmat.not.signif=6,
lwd.lmat.signif=1, lwd.lmat.not.signif=1,
lty.iso=7, col.iso="darkgray", lwd.iso=1,
lty.contr0=2, col.contr0="darkgray", lwd.contr0=1,
decdigits.ybar=2,
...
)

Arguments
x mmc.multicomp object
xlab "contrast value". An alternate "" can help unclutter a figure when sev-
eral MMC plots are displayed together.
ylab name of response variable
focus define the factor to compute contrasts of.
main, main2 main and second line of title of plot
main.method.phrase, main2.method.phrase

ry.mmc range of values on the y-axis. It is similar to par("ylim"), but not the same
as additional calculations are needed to maintain the isomeans grid as a square.
key.x, key.y location of the key displayed when iso.name=FALSE.
method method used to construct contrasts and confidence intervals. See the type ar-
gument to glht for the list.
print.lmat logical. If TRUE, then display the user-specified contrasts.
print.mca logical. If TRUE, then display the pair-wise contrasts.
iso.name logical. If TRUE, label the isomeans grid with the factor levels. If FALSE,
label the isomeans grid with sequential numbers and display a key relating the
numbers to the factor levels.
plot.mmc.multicomp 89

x.offset amount to move the vertical 0 line to the left or right to reduce overprinting of
labels and plotted lines.
col.mca.signif, lty.mca.signif, lwd.mca.signif

col.mca.not.signif, lty.mca.not.signif, lwd.mca.not.signif

col.lmat.signif, lty.lmat.signif, lwd.lmat.signif

color, line type, line width for significant user-specified contrasts.
col.lmat.not.signif, lty.lmat.not.signif, lwd.lmat.not.signif
color, line type, line width for non-significant user-specified contrasts.
lty.iso, col.iso, lwd.iso
color, line type, line width for the isomeans grid.
lty.contr0, col.contr0, lwd.contr0
color, line type, line width for the vertical contrast=0 line.
decdigits.ybar
number of decimal digits in the left-axis labels.
... other arguments, currently ignored.

Note
plot.mmc.multicomp chooses sensible defaults for its many arguments. They will often need
manual adjustment. The examples show several types of adjustments. We have changed the center-
ing and scaling to avoid overprinting of label information. By default the significant contrasts are
shown in a more intense color than the nonsignificant contrasts. We have an option to reduce the
color intensity of the isomeans grid.
When there is overprinting of labels (a consequence of level means being close together), a tiebreaker
plot may be needed. See ?MMC for an example.

Author(s)
Richard M. Heiberger <rmh@temple.edu>

## continue with the example in mmc

catalystm <- read.table(hh("datasets/catalystm.dat"), header=FALSE,
col.names=c("catalyst","concent"))
if.R(r=
oldcon <- options(contrasts = c("contr.treatment", "contr.treatment"))
,s={})

catalystm$catalyst <- factor(catalystm$catalyst, labels=c("A","B","C","D"))

catalystm1.aov <- aov(concent ~ catalyst, data=catalystm)

if.R(r=
options(oldcon)
,s={})

catalystm.mca <-
if.R(r=glht(catalystm1.aov, linfct = mcp(catalyst = "Tukey")),
s=multicomp(catalystm1.aov, plot=F))

catalystm.lmat <- cbind("AB-D" =c(0, 1, 1, 0,-2),

"A-B" =c(0, 1,-1, 0, 0),
"ABD-C"=c(0, 1, 1,-3, 1))
if.R(r=catalystm.lmat <- catalystm.lmat[-2,],
s={})
dimnames(catalystm.lmat)[[1]] <-
if.R(s=dimnames(catalystm.mca$lmat)[[1]],
r=dimnames(catalystm.mca$linfct)[[2]])

catalystm.mmc <-
if.R(r=glht.mmc(catalystm1.aov, linfct = mcp(catalyst = "Tukey"),
lmat=catalystm.lmat)
,s=multicomp.mmc(catalystm1.aov, lmat=catalystm.lmat,
plot=FALSE)
)

old.mar <- if.R(s=par(mar=c(5,12,4,6)+.1),

r=par(mar=c(5,6,4,4)+.1))

## pairwise contrasts, default settings

plot(catalystm.mmc, print.lmat=FALSE)

## centering, scaling, emphasize significant contrasts

plot(catalystm.mmc, x.offset=1.6, ry.mmc=c(50,58), print.lmat=FALSE,
col.mca.signif='red')

## user-specified contrasts
plot(catalystm.mmc, x.offset=1.6, ry.mmc=c(50,58),
col.lmat.signif='blue')

## reduce intensity of isomeans grid, number isomeans grid lines

plot(catalystm.mmc, x.offset=1.6, ry.mmc=c(50,58),
lty.iso=2, col.iso='darkgray', iso.name=FALSE,
plot.multicomp 91

col.lmat.signif='blue')

## both pairwise contrasts and user-specified contrasts

plot(catalystm.mmc, x.offset=1.6, ry.mmc=c(50,58), lty.iso=2,
col.iso='darkgray', print.mca=TRUE,
col.lmat.signif='blue', col.mca.signif='red')

par(old.mar)

plot.multicomp Multiple comparisons plot that gives independent user control over the
appearance of the significant and not significant comparisons.

Description
Multiple comparisons plot that gives independent user control over the appearance of the significant
and not significant comparisons. In R, both plot.multicomp plot.multicomp.hh coerce
their argument to an "glht" object and plots that with the appropriate plot method. In S-
Plus, plot.multicomp.hh augments the standard plot.multicomp to give additional user
arguments to control the appearance of the plot.

Usage
## S3 method for class 'multicomp':
plot(x, ...) ## R only

## S3 method for class 'multicomp.hh':

plot(x, ylabel = x$ylabel, href = 0, uniform = TRUE,
plt.in = c(0.2, 0.9, 0.1, 0.9),
x.label.adj=1,
xrange.include=href,
xlim,
comparisons.per.page=21,
col.signif=1, col.not.signif=1,
lty.signif=4, lty.not.signif=4,
lwd.signif=1, lwd.not.signif=1,
...,
xlabel.print=TRUE)

Arguments
x A "multicomp" object.
ylabel Y label on graph.
... other arguments to plot.multicomp.
href reference line for the intervals. The default is 0. S-Plus only.
xrange.include
xlim will be extended to include these values. S-Plus only.
92 plot.multicomp

uniform S-Plus only. Logical value, if TRUE and the plots fill more than one page, the
scale will be uniform across pages.
plt.in S-Plus only. Value for par("plt") to make better use of the space on the
plotting page.
x.label.adj S-Plus only. This is the par("adj") applied to the x-location of the y.labels
on the multicomp plot.
xlim x-range of the plot.
comparisons.per.page
The default S-Plus plot.multicomp hardwires this to 21, which allows for
all pairwise comparisons of 7 levels taken 2 at a time. The HH plot.multicomp
makes it a variable. Use it together with plt.in to make better use of the space
on the plot. S-Plus only.
col.signif, lty.signif, lwd.signif
Color, line type, and line width for significant comparisons. S-Plus only.
col.not.signif, lty.not.signif, lwd.not.signif
Color, line type, and line width for non-significant comparisons. S-Plus only.
xlabel.print logical. When TRUE, the caption under the plot is printed. When FALSE, the
caption under the plot is not printed. It is helpful to set this to FALSE when the
multicomp plot is used as a tiebreaker plot for the MMC plot. S-Plus only.

Value
plot.multicomp plots a "multicomp" object. In S-Plus, this masks the standard plot.multicomp
in order to provide additional arguments for controlling the appearance. It defaults to the standard
appearance. In R, it coerces its argument to an "glht" object and plots that with the appropriate
plot method.

Note
The multiple comparisons calculations in R and S-Plus use completely different libraries. Multiple
comparisons in R are based on glht. Multiple comparisons in S-Plus are based on multicomp.
The MMC plot in the HH library is the same in both systems. The details of getting the plot differ.

Author(s)
Richard M. Heiberger <rmh@temple.edu>

See Also
mmc in both languages, glht in R, multicomp in S-Plus.
position 93

Examples
## data and ANOVA
catalystm <- read.table(hh("datasets/catalystm.dat"), header=FALSE,
col.names=c("catalyst","concent"))
catalystm$catalyst <- factor(catalystm$catalyst, labels=c("A","B","C","D"))

catalystm1.aov <- aov(concent ~ catalyst, data=catalystm)

summary(catalystm1.aov)

catalystm.mca <-
if.R(r=glht(catalystm1.aov, linfct = mcp(catalyst = "Tukey")),
s=multicomp(catalystm1.aov, plot=FALSE))
if.R(s=plot(catalystm.mca),
r=plot(confint(catalystm.mca, calpha=qtukey(.95, 4, 12)/sqrt(2))))
## calpha is strongly recommended in R with a large number of levels
## See ?MMC for details.

position Find or assign the implied position for graphing the levels of a factor.
A new class "positioned", which inherits from "ordered" and "factor",
is defined.

Description
The default values for plotting a factor x are the integers 1:length(levels(x)). These func-
tions provide a way of specifying alternate plotting locations for the levels.

Usage
position(x)

position(x) <- value

## S3 method for class 'positioned':

is.numeric(x, ...)
## S3 method for class 'positioned':
as.numeric(x, ...)
## S3 method for class 'positioned':
as.double(x, ...)
## S3 method for class 'positioned':
x[..., drop=FALSE]
as.positioned(x)
is.positioned(x)
positioned(x, ..., value)
## S3 method for class 'positioned':
print(x, ...)
## S3 method for class 'positioned':
unique(x, incomparables = FALSE, ...)
unpositioned(x, ...)
94 position

Arguments
x numeric vector or factor
value numerical values to be associated with levels(x). The length(value)
must equal length(levels(as.factor(x))).
... other arguments.
drop See "["in R or "[" in S-Plus.
incomparables
See uniquein R or unique in S-Plus.

Value
position(x) <- value first forces its argument to be an ordered factor and then assigns
the value to the "position" attribute of the ordered factor. The result is assigned class
"positioned" and returned.
position(x) returns the position values associated with levels(x). If x is a positioned
factor, then the "postion" attribute is returned. If x is a factor, or an ordered factor whose levels
cannot be coerced to numeric, then the integers 1:length(levels(x)) are returned. If x is an
an ordered factor whose levels can be coerced to numeric, then the numeric values corresponding
to the levels are returned. If x is numeric, then x itself is returned.
as.numeric.positioned(x) returns a numeric vector the length of the original vector. If
x inherits from "factor", then the values in the vector are the values in position(x) sub-
scripted by the levels of the factor. If x is numeric, then x itself is returned.
unpositioned(x) removes the "position" attribute and removes the "positioned"
value from the the oldClass of the object.

Author(s)
Richard M. Heiberger <rmh@temple.edu>

See Also
panel.interaction2wt, factor in R, factor in S-Plus.

Examples
## ordered with character levels defaults to
## integer position of specified levels
tmp <- ordered(c("mm","cm","m","m","mm","cm"),
levels=c("mm","cm","m")) ## size order
tmp
as.numeric(tmp)
levels(tmp)
position(tmp)
as.numeric(tmp)

## position is assigned to ordered in specified order

tmp <- ordered(c("cm","mm","m","m","mm","cm"),
levels=c("mm","cm","m")) ## size order
position 95

levels(tmp)
position(tmp) <- c(-3, -2, 0) ## log10 assigned in size order
tmp
as.numeric(tmp)
levels(tmp)
position(tmp)
as.numeric(tmp)
unpositioned(tmp)
unique(tmp)

## numeric stays numeric

tmp <- c(0.010, 0.001, 1.000, 1.000, 0.001, 0.010)
tmp
as.numeric(tmp)
levels(tmp)
position(tmp)
as.numeric(tmp)

## factor with numeric levels, position is integer position in size order

tmp <- factor(c(0.010, 0.001, 1.000, 1.000, 0.001, 0.010))
tmp
as.numeric(tmp)
levels(tmp)
position(tmp)
as.numeric(tmp)

## ordered with numeric levels, position is numeric value in size order

tmp <- ordered(c(0.010, 0.001, 1.000, 1.000, 0.001, 0.010))
tmp
as.numeric(tmp)
levels(tmp)
position(tmp)
as.numeric(tmp)

## factor with numeric levels

## position is assigned in size order
tmp <- factor(c(0.010, 0.001, 1.000, 1.000, 0.001, 0.010))
levels(tmp)
position(tmp) <- c(-3, -2, 0) ## log10 assigned in size order
tmp
as.numeric(tmp)
levels(tmp)
position(tmp)
as.numeric(tmp)

## boxplots coded by week

tmp <- data.frame(Y=rnorm(40, rep(c(20,25,15,22), 10), 5),
week=ordered(rep(1:4, 10)))
position(tmp$week) <- c(1, 2, 4, 8)

if.R(r=
bwplot(Y ~ week, horizontal=FALSE,
96 positioned-class

scales=list(x=list(limits=c(0,9),
at=position(tmp$week),
labels=position(tmp$week))),
data=tmp, panel=panel.bwplot.intermediate.hh)
,s=
t(bwplot(week ~ Y, at=position(tmp$week),
scales=list(y=list(limits=c(0,9),
at=position(tmp$week), labels=position(tmp$week))),
data=tmp, panel=panel.bwplot.intermediate.hh))
)

#### You probably don't want to use the next two examples.
#### You need to be aware of their behavior.
##
## factor with character levels defaults to
## integer position of sorted levels.
## you probably DON'T want to do this!
tmp <- factor(c("cm","mm","m","m","mm","cm")) ## default alphabetic order
tmp
as.numeric(tmp)
levels(tmp) ## you probably DON'T want to do this!
position(tmp) ## you probably DON'T want to do this!
as.numeric(tmp)
##
## position is assigned to factor in default alphabetic order.
## you probably DON'T want to do this!
tmp <- factor(c("cm","mm","m","m","mm","cm"))
levels(tmp)
position(tmp) <- c(-3, -2, 0) ## assigned in default alphabetic order
tmp
as.numeric(tmp)
levels(tmp) ## you probably DON'T want to do this!
position(tmp) ## you probably DON'T want to do this!
as.numeric(tmp)

positioned-class Class "positioned", extends "ordered" to specify the position for

graphing the levels of a factor.

Description
The default values for plotting a factor x are the integers 1:length(levels(x)). This class
and its functions provide a way of specifying alternate plotting locations for the levels.

Objects from the Class

A virtual Class: No objects may be created from it.
print.tsdiagplot 97

Extends
Class "ordered", directly. Class "factor", by class "ordered", distance 2. Class "oldClass",
by class "ordered", distance 3.

Methods
No methods defined with class "positioned" in the signature. S3-type methods are "[.positioned",
as.double.positioned, as.numeric.positioned, as.positioned, is.numeric.positioned,
is.positioned, positioned, print.positioned, unique.positioned. Although
interaction.positioned should be a method, it isn’t because interaction is not a
generic and can’t easily be made one since the name interaction.plot conflicts.

Author(s)
Richard M. Heiberger <rmh@temple.edu>

print.tsdiagplot Print a "tsdiagplot" object.

Description
Print a "tsdiagplot" object.

Usage
## S3 method for class 'tsdiagplot':
print(x, ...)
print1.tsdiagplot(x)
print2.tsdiagplot(x)

Arguments
x a "tsdiagplot" object
... Optional arguments to print. The only argument that is used is pages.
If pages is not used or pages==1, then use print1.tsdiagplot. If
pages!=1, then use print2.tsdiagplot.
98 push.vp.hh

Details
A "tsdiagplot" object is a collection of several "trellis" objects. We provide two options
for printing them.

Author(s)
Richard M. Heiberger (rmh@temple.edu)

## linear and quadratic regressions

hardness.lin.lm <- lm(hardness ~ density, data=hardness)
hardness.quad.lm <- lm(hardness ~ density + I(density^2), data=hardness)

anova(hardness.quad.lm) ## quadratic term has very low p-value

par(mfrow=c(1,2))

regr1.plot(hardness$density, hardness$hardness,
resid.plot="square",
resid.squares 101

main="squared residuals for linear fit",

xlab="density", ylab="hardness",
points.yhat=FALSE,
xlim=c(20,95), ylim=c(0,3400))

regr1.plot(hardness$density, hardness$hardness,
model=hardness.quad.lm,
resid.plot="square",
main="squared residuals for quadratic fit",
xlab="density", ylab="hardness",
points.yhat=FALSE,
xlim=c(20,95), ylim=c(0,3400))

par(mfrow=c(1,1))

resid.squares plot squared residuals in inches to match the y-dimension

Description

plot squared residuals in inches to match the y-dimension

Usage

resid.squares(x, y, y.hat, resid.plot = "square")

Arguments

x x values
y observed y values
y.hat predicted y values
resid.plot If "square", then plot the squared residuals. If TRUE (or anything else), then
plot straight lines for the residuals.

Details

The goal is to get real squares on the screen or paper. The trick is to play games with the aspect
ratio. We find the number of inches that each vertical residual occupies. We then find the number
of x-units that corresponds to, and plot a rectangle with height=height in the y-data units and with
width=the number of x-units that we just calculated.

Author(s)

Richard M. Heiberger <rmh@temple.edu>

102 residual.plots

References

See Also

regr1.plot

Examples
hardness <- read.table(hh("datasets/hardness.dat"), header=TRUE)
hardness.lin.lm <- lm(hardness ~ density, data=hardness)

plot(hardness ~ density, data=hardness, xlim=c(22,73), ylim=c(0,3400))

abline(hardness.lin.lm)
resid.squares(hardness$density, hardness$hardness,
predict(hardness.lin.lm))

plot(hardness ~ density, data=hardness, xlim=c(22,73), ylim=c(0,3400))

abline(hardness.lin.lm)
resid.squares(hardness$density, hardness$hardness,
predict(hardness.lin.lm), resid.plot = "line")

residual.plots Residual plots for a linear model.

Description

Residual plots for a linear model. Four sets of plots are produced: (1) response against each of
the predictor variables, (2) residuals against each of the predictor variables, (3) partial residuals for
each predictor against that predictor ("partial residuals plots", and (4) partial residuals against the
residuals of each predictor regressed on the other predictors ("added variable plots").

Usage

residual.plots(lm.object, X=dft$x,
layout=c(dim(X)[2],1),
par.strip.text=list(cex=.8),
scales.cex=.6,
na.action=na.pass,
y.relation="free",
...)
residual.plots 103

Arguments
lm.object An object inheriting from "lm". It may be necessary for the lm.object to
be constructed with arguments x=TRUE, y=TRUE.
X The x matrix of predictor variables used in the linear model lm.object.
layout, par.strip.text
trellis or lattice arguments.
scales.cex cex argument forwarded to the scales argument of xyplot.
na.action A function to filter missing data. See lm.
y.relation See relation in the discussion of the scales argument in trellis.args in
S-Plus and in xyplot in R.
... Other arguments for xysplom or xyplot.

Value
A list of four trellis objects, one for each of the four sets of plots. The objects are named "y.X",
"res.X" "pres.X", "pres.Xj". The default "printing" of the result will produce four pages
of plots, one set per page. They are often easier to read when all four sets appear as separate rows
on one page (this usually requires an oversize device), or two rows are printed on each of two pages.

Author(s)
Richard M. Heiberger <rmh@temple.edu>

Examples
if.R(s={
longley <- data.frame(longley.x, Employed = longley.y)
},r={
data(longley)
})

longley.lm <- lm( Employed ~ . , data=longley, x=TRUE, y=TRUE)

## 'x=TRUE, y=TRUE' are needed to pass the S-Plus CMD check.
## They may be needed if residual.plots() is inside a nested set of
## function calls.

tmp <- residual.plots(longley.lm)

## print two rows per page

print(tmp[[1]], position=c(0, 0.5, 1, 1.0), more=TRUE)
print(tmp[[2]], position=c(0, 0.0, 1, 0.5), more=FALSE)
print(tmp[[3]], position=c(0, 0.5, 1, 1.0), more=TRUE)
print(tmp[[4]], position=c(0, 0.0, 1, 0.5), more=FALSE)
104 seqplot

seqplot Time series plot.

Description
Time series plot.

Usage
seqplot(xts, ...)

## Default S3 method:
seqplot(xts,
pch.seq=letters,
groups=as.numeric(cycle(xts)),
a=NULL, b=NULL, h=NULL, v=NULL,
ylab=deparse(substitute(xts)),
xlab=ifelse(is.null(units(xts)),"Time",units(xts)),
lwd=0, lty=c(1,3),
type="b",
...)

## S3 method for class 'rts':

seqplot(xts, pch.seq=letters, groups=as.numeric(cycle(xts)),
x.at=pretty(time(xts)[groups==min(groups)]),
x.labels,
ylab=deparse(substitute(xts)),
...)

## S3 method for class 'ts':

seqplot(xts, pch.seq=letters, groups=as.numeric(cycle(xts)),
x.at=pretty(time(xts)[groups==min(groups)]),
x.labels,
ylab=deparse(substitute(xts)),
...)

Arguments
xts Time series
pch.seq sequence of pch characters for use with the time series. The characters repeat
over the cycle of the series.
groups Numeric vector used to choose the plotting characters over cycles.
seqplot.forecast 105

a, b, h, v Arguments to panel.abline.
ylab, xlab, lwd, lty, type
stndard trellis arguments.
x.at, x.labels
shortcut for scales=list(x=list(at=x.at, labels=x.labels))
... Additional arguments to xyplot.

Author(s)
Richard M. Heiberger (rmh@temple.edu)

seqplot.forecast seqplot with confidence bands for the forecast region.

Description
seqplot with confidence bands for the forecast region.

Usage
seqplot.forecast(xts, forecast, multiplier = 1.96,
series = deparse(substitute(observed)), ylim,
CI.percent=round((1-2*(1-pnorm(multiplier)))*100,2),
main = paste(
series, " with forecast + ",
CI.percent, "% CI", sep=""),
xlab=NULL, ylab=NULL,
...) ## x.at, xlim

Arguments
xts This is the observed series
forecast forecast values based on the model
multiplier Half-width of confidence interval in standard normal units. Defaults to 1.96.
CI.percent Width of confidence band. Defaults to the standard normal, two-sided value
associated with the multiplier (95 percent for the default multiplier=1.96).
series Name of time series will be used to construct the main title for the plot.
ylim, xlab, ylab, main
standard trellis parameters
... additional arguments to xyplot.
106 strip.xysplom

Author(s)
Richard M. Heiberger (rmh@temple.edu)

sufficient Calculates the mean, standard deviation, and number of observations

in each group of a data.frame that has one continuous variable and
two factors.

Description
Calculates the mean, standard deviation, and number of observations in each group of a data.frame
that has one continuous variable and two factors.

Usage
sufficient(x,
yname = dimnames(x)[[2]][[1]],
factor.names.keep = dimnames(x)[[2]][-c(1, 2)])

Arguments
x data.frame containing a continuous variable and two factors.
yname Character name of response variable.
factor.names.keep
Character vecotr containing the names of two factors in the x data.frame.

Value
Data.frame containing five columns and as many rows as are implied by the crossing of the two
factors. Each row contains the mean in a column with the name yname and its factor values in
columns named with the name in factor.names.keep. The standard deviation of the obser-
vations in the group are in the column "sd" and the number of observations in the group is in the
column "nobs".

Author(s)
Richard M. Heiberger <rmh@temple.edu>
108 summary.arma.loop

See Also

intxplot

summary.arma.loop summary and print and subscript methods for tsdiagplot and related
objects.

Description

summary and print and subscript methods for tsdiagplot and related objects.

Usage

## S3 method for class 'arma.loop':

summary(object, ...)
## S3 method for class 'arma.loop.list':
summary(object, ...)
## S3 method for class 'arma.loop':
print(x, ...)
## S3 method for class 'arma.loop.list':
print(x, ...)
## S3 method for class 'tsacfplots':
print(x, ...)
## S3 method for class 'arma.loop':
x[..., drop = TRUE]
## S3 method for class 'diag.arma.loop':
x[..., drop = TRUE]

Arguments

x, object object to be summarized or printed or subscripted.

... additional arguments
drop See Extract in R, Subscript in S-Plus.

Author(s)

Richard M. Heiberger (rmh@temple.edu)

See Also

arma.loop, tsacfplots, tsdiagplot

t.trellis 109

t.trellis Interchange the x- and y-axes for an S-Plus trellis object. Interchange
the conditioning variables for an R trellis object.

Description
In S-Plus, change the "trellis" object to effectively, and after-the-fact convert formulas from
(y ~ x | g) to (x ~ y | g). This is needed in S-Plus because most S-Plus trellis functions
do not permit factors on the right-hand side of the formula.
In R lattice functions, factors are permitted on the right-hand side of the formula. Therefore, to
change the x and y axes within each panel, change the formula from (y ~ x | g) to (x ~
y | g). The HH t.trellis function in R calls lattice::t.trellis to interchange the
conditioning variables for an R trellis object. See t.trellis.

Usage
## S3 method for class 'trellis':
t(x)

Arguments
x any "trellis" object.

Details
In S-Plus, a warning is generated if the panel component of the "trellis" object is a func-
tion. No warning is generated when the panel component of the "trellis" object is a char-
acter string naming the function. We interchange the x and y values of the c("x", "y",
"xlab", "ylab", "xlim", "ylim", "adj.xlim", "adj.ylim") components of
the "trellis" object, and are aware of special features of the panel functions listed in the S-
Plus version of t.trellis. The transpose will usually work correctly for other panel functions.

Value
In S-Plus, a "trellis" object with all x and y components interchanged. In R, the argument
"trellis" object is sent to lattice::t.trellis to interchange the conditioning variables.

Author(s)
Richard M. Heiberger <rmh@temple.edu>

See Also
In R: t.trellis. In S-Plus: xyplot.

Examples
tmp <- data.frame(y=rnorm(30), x=factor(rep(1:3,10)))
if.R(r=
bwplot(y ~ x, data=tmp)
,s=
t(bwplot(x ~ y, data=tmp))
)

tsacfplots Coordinated time series and ACF and PCF plots.

Description
Coordinated time series and ACF and PCF plots.

Usage
tsacfplots(x,
ylab=deparse(substitute(x)),
x.name=ylab[[1]],
main=paste("Series:", x.name),
lag.at=NULL,
lag.max=NULL,
lag.units=NULL,
lag.0=TRUE,
...)

acf.pacf.plot(x,
ylab=NULL,
series=deparse(substitute(x)),
main=paste("ACF and PACF:", series),
lag.max,
lag.units=frequency(x),
lag.at=pretty(apacf$lag),
lag.labels=lag.at*lag.units,
lag.0=TRUE,
...)

Arguments
x time series
ylab, main standard trellis arguments.
tsdiagplot 111

x.name, series
Character string, name for the time series.
lag.at Location of ticks for the acf and pacf plots.
lag.labels Labels for ticks for the acf and pacf plots.
lag.max Maximum lag used in the acf and pacf plots.
lag.units Units for time series, defaults to frequency(x)
lag.0 Logical. If TRUE, then plot the correlation (identically 1) at lag=0. If FALSE,
do not plot the correlation at lag=0.
... Additional arguments to seqplot for tsacfplots. Additional arguments
to strip.default for acf.pacf.plot.

Details

The acf and pacf plots are scaled identically.

Value

"tsacfplots" object containing two "trellis" objects.

Author(s)

Richard M. Heiberger (rmh@temple.edu)

See Also

seqplot

Examples

tsacfplots(co2)
acf.pacf.plot(co2)

tsdiagplot Times series diagnostic plots for a structured set of ARIMA models.

Description

Times series diagnostic plots for a structured set of ARIMA models.

112 tsdiagplot

Usage

tsdiagplot(x,
p.max=2, q.max=p.max,
model=c(p.max, 0, q.max), ## S-Plus
order=c(p.max, 0, q.max), ## R
lag.max=36, gof.lag=lag.max,
armas=if.R(
s=arma.loop(x, model=model,
series=deparse(substitute(x)), ...),
r=arma.loop(x, order=order,
series=deparse(substitute(x)), ...)),
diags=diag.arma.loop(armas, x,
lag.max=lag.max,
gof.lag=gof.lag),
ts.diag=rearrange.diag.arma.loop(diags),
lag.units=ts.diag$tspar["frequency"],
lag.lim=range(pretty(ts.diag$acf$lag))*lag.units,
lag.x.at=pretty(ts.diag$acf$lag)*lag.units,
lag.x.labels={tmp <- lag.x.at
tmp[as.integer(tmp)!=tmp] <- ""
tmp},
lag.0=TRUE,
main, lwd=0,
...)

acfplot(rdal, type="acf",
main=paste("ACF of std.resid:", rdal$series,
" model:", rdal$model),
lag.units=rdal$tspar["frequency"],
lag.lim=range(pretty(rdal[[type]]$lag)*lag.units),
lag.x.at=pretty(rdal[[type]]$lag)*lag.units,
lag.x.labels={tmp <- lag.x.at
tmp[as.integer(tmp)!=tmp] <- ""
tmp},
lag.0=TRUE,
xlim=xlim.function(lag.lim/lag.units),
...)

aicsigplot(z, z.name=deparse(substitute(z)), series.name="ts",

model=NULL,
xlab="", ylab=z.name,
main=paste(z.name, series.name, model),
layout=c(1,2), between=list(x=1,y=1), ...)

residplot(rdal,
main=paste("std.resid:", rdal$series,
" model:", rdal$model),
...)
tsdiagplot 113

gofplot(rdal,
main=paste("P-value for gof:", rdal$series,
" model:", rdal$model),
lag.units=rdal$tspar["frequency"],
lag.lim=range(pretty(rdal$gof$lag)*lag.units),
lag.x.at=pretty(rdal$gof$lag)*lag.units,
lag.x.labels={tmp <- lag.x.at
tmp[as.integer(tmp)!=tmp] <- ""
tmp},
xlim=xlim.function(lag.lim/lag.units),
pch=16, ...)

Arguments
x Time series vector.
p.max, q.max Maximum number of AR and MA arguments to use in the series of ARIMA
models.
model A valid S-Plus model for arima.mle.
order A valid R order for arima. The additional argument seasonal may also
be used.
lag.max Maximum lag for the acf and pacf plots.
gof.lag Maximum lag for the gof plots.
armas An arma.loop object.
diags An diag.arma.loop object.
ts.diag, rdal
A list constructed as a rearranged diag.arma.loop object.
lag.units Units for time series, defaults to frequency(x)
lag.lim scaling for xlim in acf and pacf plots.
lag.x.at, lag.x.labels
Location of ticks and labels for the acf and pacf plots.
lag.0 Logical. If TRUE, then plot the correlation (identically 1) at lag=0. If FALSE,
do not plot the correlation at lag=0.
type "acf" or "pacf"
z A matrix constructed as the aic or sigma2 component of the sumamry of a
arma.loop object.
z.name "aic" or "sigma2"
series.name Character string describing the time series.
xlab, ylab, layout, between, pch, xlim, main, lwd
Standard trellis arguments.
... Additional arguments. tsdiagplot sends them to arima or arima.mle.
acfplot, aicsigplot residplot, and gofplot send them to xyplot.
114 tsdiagplot

Value

tsdiagplot returns a "tsdiagplot" object which is a list of "trellis" objects. It is

printed with its own print method.
The other functions return "trellis" objects.

Author(s)

Richard M. Heiberger (rmh@temple.edu)

References

See Also

tsacfplots, arma.loop

Examples

X <- as.rts(scan(hh("datasets/tser.mystery.X.dat")))

X.dataplot <- tsacfplots(X, lwd=1, pch.seq=16, cex=.7)

X.dataplot

X.loop <- if.R(

s=
arma.loop(X, model=list(order=c(2,0,2)))
,r=
arma.loop(X, order=c(2,0,2))
)
X.dal <- diag.arma.loop(X.loop, x=X)
X.diag <- rearrange.diag.arma.loop(X.dal)
X.diagplot <- tsdiagplot(armas=X.loop, ts.diag=X.diag, lwd=1)
X.diagplot

X.loop
X.loop[["1","1"]]
vif 115

vif Calculate the Variance Inflation Factor

Description
The VIF for predictor i is 1/(1 − Ri2 ), where Ri2 is the R2 from a regression of predictor i against
the remaining predictors.

Usage
vif(x, ...)

## Default S3 method:
vif(x, y.name, na.action = na.exclude, ...) ## x is a data.frame

## S3 method for class 'formula':

vif(x, data, na.action = na.exclude, ...) ## x is a formula

## S3 method for class 'lm':

vif(x, na.action = na.exclude, ...) ## x is a "lm" object computed with x=TRUE

Details
A simple diagnostic of collinearity is the variance inflation factor, VIF one for each regression
coefficient (other than the intercept). Since the condition of collinearity involves the predictors but
not the response, this measure is a function of the X’s but not of Y . The VIF for predictor i is
1/(1 − Ri2 ), where Ri2 is the R2 from a regression of predictor i against the remaining predictors. If
Ri2 is close to 1, this means that predictor i is well explained by a linear function of the remaining
predictors, and, therefore, the presence of predictor i in the model is redundant. Values of VIF
exceeding 5 are considered evidence of collinearity: The information carried by a predictor having
such a VIF is contained in a subset of the remaining predictors. If, however, all of a model’s
regression coefficients differ significantly from 0 (p-value < .05), a somewhat larger VIF may be
tolerable.

Value
Vector of VIF values, one for each X-variable.
116 xysplom

Author(s)
Richard M. Heiberger <rmh@temple.edu>

usair.lm <- lm(lnSO2 ~ temp + lnmfg + wind + precip, data=usair, x=TRUE)

vif(usair.lm) ## the lm object must be computed with x=TRUE

vif(lnSO2 ~ temp + lnmfg + wind + precip, data=usair)

vif(usair)

vif(usair, y.name="lnSO2")

xysplom scatterplot matrix with potentially different sets of variables on the

rows and columns.

Description
scatterplot matrix with potentially different sets of variables on the rows and columns. The slope or
regression coefficient for simple least squares regression can be displayed in the strip label for each
panel.

Usage
xysplom(x, ...)

## S3 method for class 'formula':

xysplom(x, data = sys.parent(), na.action = na.pass, ...)
xysplom 117

## Default S3 method:
xysplom(x, y=x, group, relation="free",
x.relation=relation, y.relation=relation,
xlim.in=NULL, ylim.in=NULL,
corr=FALSE, beta=FALSE, abline=corr||beta, digits=3,
x.between=NULL, y.between=NULL,
between.in=list(x=x.between, y=y.between),
scales.in=list(
x=list(relation=x.relation, alternating=FALSE),
y=list(relation=y.relation, alternating=FALSE)),
strip.in=strip.xysplom,
pch=16, cex=.75,
panel.input=panel.xysplom, ...,
cartesian=TRUE,
plot=TRUE)

Arguments
x In the "formula" method, a formula. In the "default" method, a data.frame
... other arguments to xyplot.
data data.frame
na.action See na.action in R, na.exclude in S-Plus. Defaults to na.pass because
xyplot does sensible things with missing data.
y In the "default" method, a data.frame with the same number of rows as the
data.frame in x.
group In the "default" method, a data.frame with the same number of rows as the
data.frame in x.
relation, x.relation, y.relation,scales.in
Alternate ways to get to the scales(relation=) arguments to xyplot.
xlim.in, ylim.in
Alternate ways to get to the scales(limits=) arguments to xyplot.
corr, beta Display the correlation and/or the regression coefficient for lm(y ~ x) for
each panel in an additional strip label.
abline logical. If TRUE, draw the least squares regression line within each panel. By
default the abline is FALSE unless at least one of corr or beta is TRUE.
digits number of significant digits for the correlation coefficient.
x.between, y.between, between.in
Alternate ways to get to the between= argument to xyplot.
strip.in strip function that knows how to handle the corr and beta displays.
pch, cex arguments to xyplot
panel.input panel function used by xyplot within each panel. When abline==FALSE,
the default panel function calls panel.xyplot. When abline==TRUE, the
default panel function calls panel.xyplot and panel.abline(lm(y~x,
na.action=na.exclude)). Note that we use na.action=na.exclude
inside lm.
118 xysplom

cartesian When cartesian==TRUE, the cartesian product of the left-hand side number
of variables and the right-hand side number of variables defines the number
of panels in the display. When cartesian==FALSE, each variable in the
left-hand side is paired with the variable in the corresponding position in the
right-hand side and only those pairs are plotted. Both sides must have the same
number of variables.
plot Defaults to TRUE. See details.

Details
The argument plot=TRUE is the normal setting and then the function returns a "trellis" ob-
ject. When the argument plot=FALSE, the function returns the argument list that would otherwise
be sent to xyplot. This list is interesting when the function xysplom was designed because the
function works by restructuring the input data and running xyplot on the restructured data.

Value
When plot=TRUE (the normal setting), the "trellis" object containing the graph. When
plot=FALSE, the restructured data that must be sent to the xyplot function.

Author(s)
Richard M. Heiberger <rmh@temple.edu>

## xysplom syntax options

tmp <- data.frame(y=rnorm(12), x=rnorm(12), z=rnorm(12), w=rnorm(12),

g=factor(rep(1:2,c(6,6))))
tmp2 <- tmp[,1:4]

xysplom(y ~ x , data=tmp)

xysplom(y ~ x + w , data=tmp)
xysplom(y + w ~ x , data=tmp)

xysplom(y + w ~ x , data=tmp, corr=TRUE)

xysplom(y + w ~ x , data=tmp, beta=TRUE)
xysplom(y + w ~ x , data=tmp, corr=TRUE, beta=TRUE)
xysplom 119

xysplom(y + w ~ x , data=tmp, abline=TRUE)

xysplom(y + w ~ x , data=tmp, corr=TRUE, abline=FALSE)

xysplom(y ~ x | g, data=tmp)
xysplom(y ~ x | g, data=tmp, layout=c(2,1))

xysplom(y + x ~ z | g, data=tmp)
xysplom(y + x ~ z | g, data=tmp, layout=c(2,2))
xysplom(y ~ x + z | g, data=tmp)
xysplom(y ~ x + z | g, data=tmp, layout=c(2,2))

xysplom(w + y ~ x + z, data=tmp)
xysplom(w + y ~ x + z | g, data=tmp)
xysplom(w + y ~ x + z | g, data=tmp, layout=c(2,4))

xysplom(w + y ~ x + z, data=tmp, cartesian=FALSE)

xysplom(w + y + x ~ z, data=tmp, cartesian=FALSE)
xysplom(w + y ~ x + z | g, data=tmp, cartesian=FALSE)

xysplom(w + y ~ x + z, data=tmp, scales=list(relation="same"))

xysplom(w + y ~ x + z, data=tmp, x.relation="same")

xysplom(~ y + x + z , data=tmp)
xysplom(~ y + x + z | g, data=tmp)
xysplom(~ y + x + z , data=tmp, corr=TRUE)
xysplom(~ y + x + z | g, data=tmp, corr=TRUE)
xysplom(~ y + x + z | g, data=tmp, corr=TRUE, digits=2)
xysplom(~ y + x + z | g, data=tmp, corr=TRUE, layout=c(3,6))

## Not run:
## These three examples run from R and the command line in S-Plus.
## They don't run from inside the Splus CMD check.
xysplom(~ tmp)
xysplom(~ tmp | tmp$g)
xysplom(tmp$y ~ tmp2 | tmp$g)
## End(Not run)

xysplom(g ~ x , data=tmp)
xysplom(x ~ g , data=tmp)

## Subscripting requires the x=, y= notation.

## Subscripting doesn't work with the y ~ x notation.
## Not run:
xysplom( ~ tmp[, c("x","y")]) ## doesn't work
xysplom(tmp2[, c("w","z")] ~ tmp[, c("x","y")]) ## doesn't work
## End(Not run)

## use this instead

xysplom(x = tmp[, c("x","y")])
xysplom(y = tmp2[, c("w","z")], x = tmp[, c("x","y")])

## or, even better, use the y ~ x notation

xysplom(~ x + y, data=tmp)
120 xysplom

xysplom(w + z ~ x + y, data=cbind(tmp, tmp2))

## xyplot in R has many similar capabilities with xysplom

if.R(r=
xyplot(w + z ~ x + y, data=tmp, outer=TRUE)
,s=
{}
)
Index

∗Topic NA panel.bwplott, 75
diag.maybe.null, 24 panel.ci.plot, 78
∗Topic algebra panel.dotplot.tb, 79
orthog.complete, 71 panel.pairs.hh, 82
∗Topic aplot panel.xysplom, 83
norm.curve, 63 plot.multicomp, 90
panel.interaction2wt, 80 position, 92
∗Topic classes push.vp.hh, 97
ancova-class, 7 strip.background0, 105
positioned-class, 95 strip.xysplom, 105
∗Topic color sufficient, 106
col.hh, 23 t.trellis, 108
∗Topic data ∗Topic environment
HH-package, 2 hh, 29
∗Topic design ∗Topic hplot
HH-package, 2 ae.dotplot, 5
interaction2wt, 34 ancova, 8
mmc, 49 arma.loop, 16
panel.interaction2wt, 80 ci.plot, 22
∗Topic device extra, 26
export.eps, 26 HH-package, 2
GSremove, 1 interaction2wt, 34
∗Topic distribution ladder, 40
norm.curve, 63 lm.case, 45
∗Topic dplot mmc, 49
ancova, 8 mmc.mean, 56
as.multicomp, 17 norm.curve, 63
axis.i2wt, 20 odds.ratio, 69
bwplot.formula, 20 panel.cartesian, 76
grid.yaxis.hh, 28 partial.corr, 84
interaction.positioned, 33 plot.hov, 85
intxplot, 37 plot.mmc.multicomp, 86
ladder, 40 print.tsdiagplot, 96
legendGrob2wt, 44 residual.plots, 101
multicomp.order, 60 seqplot, 103
multicomp.reverse, 62 seqplot.forecast, 104
panel.acf, 73 tsacfplots, 109
panel.bwplot.intermediate.hh, tsdiagplot, 110
74 xysplom, 115

121
122 INDEX

∗Topic htest [.arma.loop (summary.arma.loop),

ae.dotplot, 5 107
aov.sufficient, 13 [.diag.arma.loop
HH-package, 2 (summary.arma.loop), 107
mcalinfct, 48 [.mmc.multicomp (mmc), 49
mmc, 49 [.positioned (position), 92
mmc.mean, 56
odds.ratio, 69 acf.pacf.plot (tsacfplots), 109
∗Topic manip acfplot (tsdiagplot), 110
ladder, 40 AE.dotplot (ae.dotplot), 5
∗Topic models ae.dotplot, 5
ancova, 8 aicsigplot (tsdiagplot), 110
anova.mean, 11 analysis of covariance (ancova), 8
do.formula.trellis.xysplom, ancova, 3, 8, 8
25 ancova-class, 11
hov, 30 ancova-class, 7
plot.hov, 85 anova.ancova (ancova), 8
regr1.plot, 98 anova.lm, 12
resid.squares, 100 anova.mean, 11
∗Topic package aov, 11, 13, 31, 86
HH-package, 2 aov.sufficient, 13
∗Topic print arima, 16, 112
arima.diag, 15, 28
summary.arma.loop, 107
arima.diag.hh, 15
∗Topic regression
arima.mle, 16, 112
ancova, 8
arima.model (extra), 26
ci.plot, 22
arma.loop, 16, 27, 107, 113
interaction.positioned, 33
as.character.arima.model (extra),
lm.case, 45
26
regr1.plot, 98
as.double.positioned (position),
resid.squares, 100
92
residual.plots, 101
as.glht, 61
vif, 114
as.glht (as.multicomp), 17
X.residuals, 3
as.multicomp, 17, 52
∗Topic ts as.numeric.positioned (position),
arima.diag.hh, 15 92
arma.loop, 16 as.positioned (position), 92
extra, 26 as.rts (extra), 26
gof.calculation, 27 axis.default, 20
HH-package, 2 axis.i2wt, 20
npar.arma, 67
∗Topic univar bwplot, 35, 74, 75
sufficient, 106 bwplot.formula, 20, 21
∗Topic utilities
hh, 29 ci.plot, 22, 79
if.R, 32 coef.ancova (ancova), 8
objip, 68 coef.arima.HH (extra), 26
.arima.info.names.not.ordered coefficients.ancova (ancova), 8
(extra), 26 col.hh, 23
INDEX 123

compute.packet.new ladder, 40, 77

(bwplot.formula), 20 ladder3 (ladder), 40
confint, 50 legendGrob2wt, 44
confint.glht, 19, 51 lm, 4, 13, 23, 79, 115
confint.glht.hh, 19 lm.case, 4, 45
contrMat, 48 lm.influence, 47
covariance (ancova), 8 logrelrisk (ae.dotplot), 5

dev.copy2eps, 26 make.xaxis.hh.labels
diag, 24 (grid.yaxis.hh), 28
diag.arma.loop (arma.loop), 16 make.yaxis.hh.labels
diag.maybe.null, 24 (grid.yaxis.hh), 28
do.formula.trellis.xysplom, 25 mcalinfct, 48
dotplot, 79 mcp, 50
MMC, 13, 19, 48, 61, 63, 72
export.eps, 26 MMC (mmc), 49
export.graph, 26 mmc, 18, 20, 49, 58, 88, 91
extra, 26 mmc.mean, 56
Extract, 107 model.frame.ancova (ancova), 8
multicomp, 19, 20, 51, 58, 91
factor, 33, 93 multicomp (mmc), 49
formula, 25 multicomp.label.change
(multicomp.order), 60
glht, 19, 20, 48, 50, 51, 58, 87, 91 multicomp.label.change.mmc.multicomp
glht.mmc (mmc), 49 (multicomp.order), 60
gof.calculation, 27 multicomp.label.change.multicomp
gofplot (tsdiagplot), 110 (multicomp.order), 60
graphsheet, 2 multicomp.mean (mmc.mean), 56
grid.text, 27 multicomp.mmc.mean (mmc.mean), 56
grid.xaxis.hh (grid.yaxis.hh), 28 multicomp.order, 60, 63
grid.yaxis.hh, 28 multicomp.reverse, 61, 62
GSremove, 1
na.action, 4, 25, 114, 116
HH (HH-package), 2 na.exclude, 116
hh, 29 norm.curve, 63
HH-package, 2 norm.observed (norm.curve), 63
HH.ROOT.DIR (hh), 29 norm.setup (norm.curve), 63
hov, 30, 86 npar.arma, 67
npar.rarma (npar.arma), 67
if.R, 32 npar.sarma (npar.arma), 67
interaction.positioned, 33, 82
interaction2wt, 20, 34, 44, 74, 82 objects, 69
intxplot, 37, 107 objip, 68
is.numeric.positioned (position), odds.ratio, 69
92 orthog.complete, 71
is.positioned (position), 92 orthog.construct
is.R, 32 (orthog.complete), 71

key, 6 panel.acf, 73
124 INDEX

panel.ae.dotplot (ae.dotplot), 5 print1.tsdiagplot

panel.ae.leftplot (ae.dotplot), 5 (print.tsdiagplot), 96
panel.ae.rightplot (ae.dotplot), 5 print2.tsdiagplot
panel.ancova (ancova), 8 (print.tsdiagplot), 96
panel.bwplot, 75 push.vp.hh, 97
panel.bwplot.intermediate.hh, 74,
rearrange.diag.arma.loop
82
(arma.loop), 16
panel.bwplott, 75
regr1.plot, 98, 101
panel.cartesian, 43, 76, 97
resid.squares, 99, 100
panel.case (lm.case), 45
residplot (tsdiagplot), 110
panel.ci.plot, 78
residual.plots, 101
panel.dotplot.tb, 79
panel.gof (panel.acf), 73 seqplot, 103, 105, 110
panel.hov (plot.hov), 85 seqplot.forecast, 104
panel.interaction2wt, 35, 36, 80, 93 splom, 83
panel.intxplot (intxplot), 37 strip.background0, 105
panel.pairs.hh, 82 strip.default, 42, 82
panel.std.resid (panel.acf), 73 strip.interaction2wt
panel.xyplot, 10, 74 (panel.interaction2wt), 80
panel.xysplom, 83 strip.ladder (ladder), 40
partial.corr, 84 strip.xysplom, 105
plot.ancova (ancova), 8 stripplot, 79
plot.case (lm.case), 45 Subscript, 107
plot.hov, 31, 85 sufficient, 38, 39, 106
plot.mmc.multicomp, 12, 52, 58, 86 summary.ancova (ancova), 8
plot.multicomp, 90 summary.arma.loop, 107
plot.odds.ratio (odds.ratio), 69 t.trellis, 108, 108, 109
pop.vp.hh (push.vp.hh), 97 title.grob (extra), 26
position, 92, 96 title.trellis (extra), 26
position<- (position), 92 trellis.args, 10, 102
positioned, 33 trellis.device, 23
positioned (position), 92 trellis.par.get, 23
positioned-class, 95 tsacfplots, 104, 107, 109, 113
predict.ancova (ancova), 8 tsdiagplot, 15, 17, 74, 97, 107, 110
predict.lm, 23
print.ancova (ancova), 8 unique, 93
print.arma.loop unique.positioned (position), 92
(summary.arma.loop), 107 unit, 97
print.glht.mmc.multicomp units.ts (extra), 26
(as.multicomp), 17 unpositioned (position), 92
print.mmc.multicomp vcov.sufficient (aov.sufficient),
(as.multicomp), 17 13
print.multicomp (as.multicomp), 17 viewport, 97
print.positioned (position), 92 vif, 4, 114
print.trellis, 6
print.tsacfplots X.residuals, 3
(summary.arma.loop), 107 xyplot, 6, 10, 11, 79, 102, 109, 117
print.tsdiagplot, 96 xysplom, 77, 84, 106, 115

NSTP 2 Module 1 - Philippine Demography 1
No ratings yet
NSTP 2 Module 1 - Philippine Demography 1
53 pages
Trends Networks Module 4
No ratings yet
Trends Networks Module 4
34 pages
Mess PDF
100% (1)
Mess PDF
94 pages
Caret
No ratings yet
Caret
222 pages
Me Trology
No ratings yet
Me Trology
128 pages
GCDkit Manual
No ratings yet
GCDkit Manual
342 pages
MGCV
No ratings yet
MGCV
338 pages
Modern Analysis of Biological Data Generalized Linear Models in R Stano Pekar Download
No ratings yet
Modern Analysis of Biological Data Generalized Linear Models in R Stano Pekar Download
81 pages
Igraph
No ratings yet
Igraph
426 pages
The Old Man and The Sea by Ernest Hemingway: Plot Summary
No ratings yet
The Old Man and The Sea by Ernest Hemingway: Plot Summary
7 pages
AgroReg 1.2.7
No ratings yet
AgroReg 1.2.7
110 pages
Librería Fda R
No ratings yet
Librería Fda R
292 pages
Analogue
No ratings yet
Analogue
156 pages
Package HH': R Topics Documented
No ratings yet
Package HH': R Topics Documented
292 pages
Catálogo Bombas K3V y K5V
100% (3)
Catálogo Bombas K3V y K5V
15 pages
Hmisc
100% (1)
Hmisc
452 pages
GREPA Digest
No ratings yet
GREPA Digest
6 pages
Aerb CT
No ratings yet
Aerb CT
6 pages
Mirt
No ratings yet
Mirt
200 pages
The R Development Core Team-R Reference Manual - Base Package, Vol. 2-Network Theory Ltd. (2003) PDF
No ratings yet
The R Development Core Team-R Reference Manual - Base Package, Vol. 2-Network Theory Ltd. (2003) PDF
704 pages
Position Paper (Leg Writ.)
100% (1)
Position Paper (Leg Writ.)
6 pages
Package E1071': September 16, 2024
No ratings yet
Package E1071': September 16, 2024
67 pages
HH
100% (1)
HH
196 pages
Package Car': September 27, 2024
No ratings yet
Package Car': September 27, 2024
160 pages
Instruction Manual Single Method - M100 - Chlorine T - en PDF
No ratings yet
Instruction Manual Single Method - M100 - Chlorine T - en PDF
11 pages
Package Hmisc' - Harrell (2022)
No ratings yet
Package Hmisc' - Harrell (2022)
455 pages
Boom
No ratings yet
Boom
63 pages
Package E1071': R Topics Documented
No ratings yet
Package E1071': R Topics Documented
63 pages
Metan
No ratings yet
Metan
278 pages
Spatial Eco
100% (1)
Spatial Eco
119 pages
Assignment Activity Unit 5
No ratings yet
Assignment Activity Unit 5
6 pages
Bipartite PDF
No ratings yet
Bipartite PDF
150 pages
Smacof
No ratings yet
Smacof
78 pages
Mi̇rt Manual
No ratings yet
Mi̇rt Manual
170 pages
PETA Letter-About-Lothrop
No ratings yet
PETA Letter-About-Lothrop
5 pages
Manual de Paquete (FDA)
No ratings yet
Manual de Paquete (FDA)
293 pages
Sanket Din Rail Series New
No ratings yet
Sanket Din Rail Series New
1 page
Package MFD': February 26, 2024
No ratings yet
Package MFD': February 26, 2024
93 pages
Package E1071': R Topics Documented
No ratings yet
Package E1071': R Topics Documented
67 pages
Mice
No ratings yet
Mice
142 pages
Package Geophys': R Topics Documented
No ratings yet
Package Geophys': R Topics Documented
77 pages
Objective 7.01 Key Terms
No ratings yet
Objective 7.01 Key Terms
3 pages
Int. J. Production Economics: Mohd Helmi Ali, Norhidayah Suleiman
No ratings yet
Int. J. Production Economics: Mohd Helmi Ali, Norhidayah Suleiman
12 pages
Plotrix
No ratings yet
Plotrix
231 pages
PEX - Installation Simulator
No ratings yet
PEX - Installation Simulator
17 pages
Package Tsdyn': R Topics Documented
No ratings yet
Package Tsdyn': R Topics Documented
93 pages
Lme4 - Morfometria
No ratings yet
Lme4 - Morfometria
113 pages
Ggbio
No ratings yet
Ggbio
266 pages
Useful
No ratings yet
Useful
60 pages
Presentation Schedule 5nov 1page
No ratings yet
Presentation Schedule 5nov 1page
6 pages
Package Dae': R Topics Documented
No ratings yet
Package Dae': R Topics Documented
129 pages
Dynamo Dynamics
No ratings yet
Dynamo Dynamics
11 pages
Random Fields
No ratings yet
Random Fields
374 pages
Tseries
No ratings yet
Tseries
54 pages
Package Seqhmm': June 17, 2019
No ratings yet
Package Seqhmm': June 17, 2019
81 pages
Mirt PDF
No ratings yet
Mirt PDF
183 pages
Reiji Ecstasy 1
No ratings yet
Reiji Ecstasy 1
5 pages
The Sna Package: Topics Documented
No ratings yet
The Sna Package: Topics Documented
187 pages
RGR PDF
No ratings yet
RGR PDF
209 pages
Dolan & John (2019) Market Intelligence. HBP
0% (1)
Dolan & John (2019) Market Intelligence. HBP
6 pages
Caret PDF
No ratings yet
Caret PDF
223 pages
Car PDF
No ratings yet
Car PDF
147 pages
FDA
No ratings yet
FDA
250 pages
Caret
0% (1)
Caret
206 pages
Car PDF
No ratings yet
Car PDF
151 pages
Package Tseries': R Topics Documented
No ratings yet
Package Tseries': R Topics Documented
54 pages
Plotrix
No ratings yet
Plotrix
231 pages
Package Caret': R Topics Documented
No ratings yet
Package Caret': R Topics Documented
136 pages
R E1071 Package Definition
No ratings yet
R E1071 Package Definition
63 pages
Lattice - Trellis Graphics For R
No ratings yet
Lattice - Trellis Graphics For R
159 pages
C18 Marine Engine-Maintenance Intervals
73% (11)
C18 Marine Engine-Maintenance Intervals
55 pages
International Structured Collateral Finance v2
No ratings yet
International Structured Collateral Finance v2
8 pages
HH PDF
No ratings yet
HH PDF
276 pages
Children Who Kill Their Parents 1984
No ratings yet
Children Who Kill Their Parents 1984
82 pages
WFA Hose Fittings PDF
No ratings yet
WFA Hose Fittings PDF
46 pages
Class 12 Chapter 10 Haloalkanes and Haloarenes-1
No ratings yet
Class 12 Chapter 10 Haloalkanes and Haloarenes-1
75 pages
GCDkit Manual PDF
No ratings yet
GCDkit Manual PDF
282 pages
SVM in R
No ratings yet
SVM in R
62 pages
GCDkit Manual
No ratings yet
GCDkit Manual
175 pages
Sejarah Dan Perkembangan Fitoterapi
No ratings yet
Sejarah Dan Perkembangan Fitoterapi
3 pages
ManualeUtente EMD031-121 en
No ratings yet
ManualeUtente EMD031-121 en
42 pages
Chip Ferrite Beads: Features
No ratings yet
Chip Ferrite Beads: Features
27 pages
The Home Health Quality Reporting Program (HH QRP) : OASIS Requirements
No ratings yet
The Home Health Quality Reporting Program (HH QRP) : OASIS Requirements
4 pages
DG Set
No ratings yet
DG Set
19 pages
GCDkit Manual
No ratings yet
GCDkit Manual
272 pages
Arab World English Journal (AWEJ) Volume 9. Number 3. September 2018 Pp.176-187
No ratings yet
Arab World English Journal (AWEJ) Volume 9. Number 3. September 2018 Pp.176-187
12 pages
Community Engagement Solidarity and Citizenship
No ratings yet
Community Engagement Solidarity and Citizenship
7 pages
Antidiabetic Property of Klorostrep
No ratings yet
Antidiabetic Property of Klorostrep
9 pages
Unmarked R Package
No ratings yet
Unmarked R Package
89 pages
Water Drop Function Test 4-8-19
No ratings yet
Water Drop Function Test 4-8-19
9 pages
Fortigate 80f Series Network Security Firewall Manual
No ratings yet
Fortigate 80f Series Network Security Firewall Manual
9 pages
Waveslim PDF
No ratings yet
Waveslim PDF
81 pages
Foul Sewer Design Report
No ratings yet
Foul Sewer Design Report
7 pages
Knit Soxx for Everyone: 25 Colorful Sock Patterns for the Whole Family
From Everand
Knit Soxx for Everyone: 25 Colorful Sock Patterns for the Whole Family
Kerstin Balke
4.5/5 (2)

Note: This service is not intended for secure transactions such as banking, social media, email, or purchasing. Use at your own risk. We assume no liability whatsoever for broken pages.