SAS Global Forum 2009 Reporting and Information Visualization

Paper 216-2009

Using SAS® Software to Generate Textbook-Style Histograms

Perry Watts, SAS® User, Elkins Park, PA

ABSTRACT
Percent tallies associated with midpoint labeled intervals define the basic histogram generated in PROC
UNIVARIATE. However, most statistics textbooks display histograms with frequencies and endpoints rather than per-
cents and midpoints. Frequencies are more descriptive, and endpoints are better suited for continuous data. With
recent updates to SAS software it is now easy to generate a textbook histogram by using PROC UNIVARIATE. En-
hancements to the textbook histogram such as a normal curve overlay and bar height labels are also easily managed
in PROC UNIVARIATE.

Unfortunately, the UNIVARIATE ENDPOINTS= option, new for Version 9.13 SAS, is restricted in form to <m TO n BY
increment>. This means that plotting an n-bar histogram or a histogram with unequal intervals is only possible when
the graph is developed from scratch in PROC GPLOT. A macro that works with any release of SAS software is pro-
vided that automates the production of GPLOT generated histograms.

With complete instructions provided for both UNIVARIATE and GPLOT derived histograms, you should come away
from this presentation knowing how to create a textbook histogram in SAS.

THE HISTOGRAM: FOR CONTINUOUS DATA

DEFINITION:
From Wikipedia:
In statistics, a histogram is a graphical display of tabulated frequencies. A histogram is the graphical version of a table which
shows what proportion of cases fall into each of several or many specified categories. The histogram differs from a bar chart in
that it is the area of the bar that denotes the value, not the height, a crucial distinction when the categories are not of uniform
width (Lancaster, 1974). The categories are usually specified as non-overlapping intervals of some variable. The categories
(bars) must be adjacent.
The word histogram is derived from Greek: histos 'anything set upright' (as the masts of a ship, the bar of a loom, or the vertical
bars of a histogram); gramma 'drawing, record, writing' (some Italics added) [5].

For the histogram, then, the width of the bar becomes an added dimension for conveying information. This means
that bar widths in a histogram do not have to be equal.
THE SAS-STYLE HISTOGRAM IS DERIVED FROM THE GCHART PROCEDURE
Prior to Version 8, the only way to quickly generate a histogram in SAS was to remove the DISCRETE option from an
invocation of PROC GCHART while setting SPACE (between bars) to zero. Just like a regular bar chart, measure-
ment classes were labeled at midpoints along the horizontal axis. Now, as Figure 1 demonstrates, the output is al-
most identical when defaults are applied to a PROC UNIVARIATE generated histogram.

Figure 1. GCHART and UNIVARIATE generate similar histograms. For GCHART, FREQUENCY is the default setting
whereas PERCENTS are plotted in UNIVARIATE. Midpoints are also marked by ticks in the UNIVARIATE histogram.

Histogram for Meeting Lengths Histogram for Meeting Lengths

(PROC GCHART) (PROC UNIVARIATE)
10 30

25
Number of Meetings

8
20
Percent

6
15
4
10

2 5

0
0 0 0.8 1.6 2.4 3.2 4
0.0 0.8 1.6 2.4 3.2 4.0
Hours Hours

1
SAS Global Forum 2009 Reporting and Information Visualization

The MEETINGS data set graphed in Figure 1 comes from The How-To Book for SAS/GRAPH Software by Thomas
Miron [2, p.88] (copyright 1995, SAS Institute Inc., Cary, NC, USA. All Rights Reserved; reproduced with permission of SAS Insti-
tute Inc., Cary, NC). Since the variation in meeting lengths is not infinite, HOURS in Figure 1 would be better described
as a discrete, continuous variable. Lots of ties exist in the small, 32-observation data set. As Figure 2 demonstrates,
the needle plot is the graph of choice for discrete continuous data whereas the histogram should be reserved for data
that are truly continuous.

Figure 2. The MEETINGS data set is graphed as a needle plot. Discrete continuous data do not need to be summarized. On the
other hand, information about 300+ baseball players can be collapsed into six bar areas of a histogram, because it is possible for
a player to score any number of runs during a given season.

Count of Meetings by their Length Baseball Data from 1986-1987

8
125

6
100
Number of Meetings

Number of Players
75
4

50

2
25

0
0
0.25 0.50 1.00 1.50 2.00 2.50 3.00 3.50 4.0
0 25 50 75 100 125 150
Hours Runs

Even though the MEETINGS data should not be plotted as a histogram, the small data set highlights structural issues
associated plot construction. Therefore, MEETINGS histograms appear alongside their BASEBALL counterparts
throughout the paper.

HISTOGRAMS FROM PROC UNIVARIATE

DEFAULT SETTINGS:
The second graph in Figure 1 is reproduced along with associated with source code in Figure 3 below. When
MIDPOINTS= or ENDPOINTS= options are not specified, a histogram is generated with midpoint values calculated
internally by SAS [8, p.225].

Figure 3. Default settings for histograms created in PROC UNIVARIATE.

Histogram for Meeting Lengths filename histog "&outpath\Fig1b.cgm";
(PROC UNIVARIATE) goptions device=cgmOf97L ctext=black
30 rotate=landscape gunit=pct
polygonfill fill ftext=HWCGM001
25 htext=4pct htitle=5pct gsfname=histog;

20 proc univariate data=histo.meetings noprint;

Percent

histogram hours / cfill=grayCC

15 noframe;
Label Hours='Hours';
10 run;

0
? 0 0.8 1.6 2.4 3.2 4
Hours

• ftext= htext= While GCHART-generated histograms can reference AXES statements, PROC
UNIVARIATE must rely on a limited set of options in the GOPTIONS statement to format values
and labels along the horizontal axis.

2
SAS Global Forum 2009 Reporting and Information Visualization

• noprint suppresses summary statistics

• cfill= in the histogram statement assigns a color to the histogram.

The internal algorithm developed by Terrel and Scott that SAS uses for default midpoint assignments is also "primar-
ily applicable to continuous data that are approximately normally distributed" [8, p.225]. Thus the MEETINGS data set
presents difficulties when defaults are used in histogram construction. Meeting lengths are not normally distributed
and the negative time as implied by the question mark in Figure 2 doesn't exist.
CONSTRUCTING A TEXTBOOK HISTOGRAM:
The results from a poll of eleven statistics references can be found in Tables 1 and 2 in the appendix. From Table 1,
nine out of the eleven references contain at least one histogram with endpoints rather than midpoints. Two midpoint
histograms from the surveyed texts present problems that are illustrated in Figure 4 below.
The first graph in Figure 4 shows how a histogram was used for quality control [12, 229-231]. The diameters of 500
rods were grouped at 0.001 intervals. The lower specification limit (LSL) was set to 1.000. If a rod diameter was less
than 1.000, the rod had to be discarded. Rods with diameters greater than 1.000 could be retooled. A zero at 0.999
raised questions about the results, and the mystery was cleared up when inspectors revealed that they passed rods
that were slightly below the lower specification limit. However, given that rod diameters could be anywhere from
0.996 to 1.008 cm inclusive in diameter, what about rods between 0.9995 and 0.9999 cm? Shouldn't they have been
rejected too? If endpoints had been used instead of midpoints, the results would have been unambiguous.
The second graph of theoretical proportions is not so convoluted [19, 171-172]. Negative proportions don't exist. Again
zero appears as (an unlabeled) midpoint, so half of the corresponding bar is out of bounds.

Figure 4. Two histograms from ten statistical references contain ambiguities caused by using midpoints rather than endpoints to
define class intervals. The histograms were re-created by visual inspection. Therefore, bar heights are estimated.

Distribution of Inside Diameters for 500 Steel Rods Distribution of a Theoretical Proportion
from 'Statistics: A Guide to the Unknown', p.230 from 'Statistics: Concepts and Controversies', p.162
100 LSL
0.2

75
Relative Frequency
Frequency

50 0.1

25

0 0 0
0 0.0
0.996 0.998 1.000 1.002 1.004 1.006 1.008 -0.04 0.04 0.12 0.20 0.28 0.36 0.44
Diameter(CM) Value of P-Hat

SWITCHING FROM MIDPOINTS TO ENDPOINTS IN VERSION 9.13 SAS:

In version 9.13, the ENDPOINTS= option was added to the histogram statement in PROC UNIVARIATE. A
UNIVARIATE endpoints histogram is displayed in Figure 5 along with the associated source code. Another endpoint s
histogram can be found in Example 3.18 Binning a Histogram from Base SAS 9.13 Procedures guide Volume 3 [8, p.
341].

By default, SAS uses a BEST format with varied precision to label class intervals in a histogram. If uniform precision
is desired, a specific format should be defined. For example, to display 0.0, 0.8, 1.5, 2.4, 3.2, 4.0, and
4.8 in Figure 5, insert "format hours 3.1;" into the code.

3
SAS Global Forum 2009 Reporting and Information Visualization

Figure 5. An endpoint histogram is created in PROC UNIVARIATE with the ENDPOINTS= option.
Histogram for Meeting Lengths filename histo "&outpath.\Fig5a.cgm";
25 goptions htext=4pct htitle=5pct
gsfname=histo;

20 proc univariate data=histo.meetings noprint;

histogram hours / cfill=ltgray
endpoints=0 to 4.8 by 0.8
15
Percent

noframe;
Label Hours='Hours';
10 run;

0
0 0.8 1.6 2.4 3.2 4 4.8
Hours

CHANGING THE Y-AXIS FROM PERCENTS TO FREQUENCIES:

In the Wikipedia definition, the vertical axis of a histogram is restricted to a display of frequencies, and from Table 2 in
the Appendix, ten out of the eleven references contain at least one frequency histogram. In contrast, the single fre-
quency histogram displayed in the Version 8 manual [8, p.1452] has been removed from the Version 9.13 manual.
Even though frequency histograms are not emphasized in the documentation, it is easy to change the percent histo-
gram in Figure 5 to a frequency histogram displayed in Figure 6. Note that the relative heights of the bars are the
same in both histograms.

Figure 6. Percents are changed to frequencies by setting the VSCALE= option to COUNT.
Histogram for Meeting Lengths filename histo "&outpath.\Fig6a.cgm";
8 goptions htext=4pct htitle=5pct
gsfname=histo;

proc univariate data=histo.meetings noprint;

6
Number of Meetings

histogram hours / cfill=ltgray

endpoints=0 to 4.8 by 0.8 noframe
vscale=count vaxis=0 2 4 6 8
4 vaxisLabel="Number of Meetings";
Label Hours='Hours';
run;
2

0
0 0.8 1.6 2.4 3.2 4 4.8
Hours

• vscale=count Other choices available are PERCENT, the default, and PROPORTION (or
relative frequency).
• vaxis= VALUE LIST In Version 9.13, a NAME associated with an axis statement can also be
used. There is no corresponding HAXIS= option for PROC UNIVARIATE.
• vaxisLabel= labels the vertical axis.

Now that frequencies can be plotted along the vertical axis of a histogram with PROC UNIVARIATE, it would be de-
sirable to attach counts to the individual bars. While this task can be easily completed in PROC GCHART with the
OUTSIDE= or INSIDE= options, ANNOTATE must be used in PROC UNIVARIATE. A labeled histogram along with
relevant SAS code is displayed in Figure 7.

4
SAS Global Forum 2009 Reporting and Information Visualization

Figure 7. Bars are labeled with output from an ANNOTATE data set that is linked to the HISTOGRAM statement in PROC
UNIVARIATE with the ANNOTATE= option.

Histogram for Meeting Lengths ™ Creating the ANNOTATE Data Set š

(PROC UNIVARIATE)
proc sql noprint;
10 select n(hours) into :tot_N from
8 8
histo.meetings;
8 quit;
Number of Meetings

proc univariate data=histo.meetings noprint;

6 histogram hours / noplot
6
endpoints=0.0 to 4.80 by 0.8
4 4 outhistogram=histoMtgsDS;
4 run;
data annoMtgsMP;
2
2 %dclanno;
length text chn $8;
%system(2,2,3);
0 set histoMtgsDS;
0 0.8 1.6 2.4 3.2 4 4.8
x = _minPT_ + 0.4;
Hours
y = &tot_n * _obsPct_ * 0.01;
chN = Left(put(Y,8.));
%label(x,y,chN,CX0386BE,0,0,3.5,HWCGM001,2);
Invoking ANNOTATE › run;
filename histo "&outpath.\Fig7a.cgm";
goptions htext=4pct htitle=5pct A Printout of the ANNOTATE Data Set œ
gsfname=histo;

proc univariate data=histo.meetings no- _VAR_ X Y _MINPT_ _OBSPCT_

print;
histogram hours / annotate=annoMtgsMP hours 0.4 8 0.0 25.00
cfill=ltgray
hours 1.2 6 0.8 18.75
endpoints=0 to 4.8 by 0.8 noframe
vscale=count vaxis=0 2 4 6 8 10 hours 2.0 4 1.6 12.50
vaxisLabel="Number of Meetings"; hours 2.8 8 2.4 25.00
Label Hours='Hours'; hours 3.6 4 3.2 12.50
run; hours 4.4 2 4.0 6.25
_MINPT_ = "Left endpoint of Histogram Interval."
_OBSPCT_ = "Percent Obs in Interval."

Panel 2:
• tot_N The total number of observed values for hours is stored in a macro variable, so that per-
cents in _OBSPCT_ can be converted to frequencies for display.
• noplot ... outhistogram= histoMtgsDS suppresses the histogram, since only the output data
set is desired in this step.
• endpoints=0.0 to 4.80 by 0.8 is needed for calculating _MINPT_ and _OBSPCT_ in the
ANNOTATE data set. . Adding vscale=count vaxis=0 2 4 6 8 10 from panel 3 to panel 2 will
not change the contents of the output data set. In other words, _OBSPCT_ is fixed. A corre-
sponding automatic variable such as _OBSFREQ_ does not exist.
• data annoMtgsMP ... set histoDS the output data set from the first invocation of PROC
UNIVARIATE is used as input to ANNOTATE.
• x = _minPT_ + 0.4; Since frequency labels are centered over the bar midpoint, one-half of the
range (0.8) or 0.4 is added to _MINPT_.
• y = &tot_n * _obsPct_ * 0.01; _OBSPCT_ is converted to a frequency.
• %label(x,y,chN,CX0386BE,0,0,3.5,HWCGM001,2) For a description of the %Label annotate
macro see [7, p. 685].
Panel 3:
• annotate=annoMtgsMP is the link to the ANNOTATE data set created in panel 2.
• endpoints=0 to 4.8 by 0.8 must contain the same range as the ENDPOINTS= option in panel
2.
• vaxis=0 2 4 6 8 10 the vertical axis maximum is increased to 10 to accommodate the labels.

5
SAS Global Forum 2009 Reporting and Information Visualization

ADDING A NORMAL CURVE AND LEGEND TO A HISTOGRAM IN PROC UNIVARIATE:

In Figure 8 normal curves with associated legends are added to three histograms. The first histogram is a repeat of
the one displayed in Figure 7, and the histograms in panels 3 and 4 are derived from larger data sets. A complete list
of NORMAL secondary options can be found in Table 3.3 on page 214 in the Version 9 Procedures guide [9],
whereas INSET options associated with the legend are described in Table 3.14 on page 230. For additional examples
of histograms with normal curves see Nguyen [1] and the Procedures Guide [9, 343-346].

Figure 8. Normal curves can be easily added to a UNIVARIATE-generated histogram. The histograms become more informative
when sample sizes and probabilities are also listed. Despite the larger sample size, the curve in panel 3 does not come from a
normal distribution whereas the curve in the panel 4 histogram is destined to be normal. The idea for the panel 4 histogram origi-
nates in Example 8 of the Version 8 Procedures Guide [8, p.1444-1446].

Histogram for Meeting Lengths ™ Adding a Normal Curve and Legend š

10 to the Histogram in Panel 1

8 8 proc univariate data=histo.meetings

8 normal noprint;
Number of Meetings

histogram hours /
6 normal(noprint color=CX0386BE w=3)
6
annotate= annoMtgsMP
cfill=ltgray
4 4
4 endpoints=0 to 4.8 by 0.8 noframe
vscale=count vaxis=0 2 4 6 8 10
2 vaxisLabel="Number of Meetings";
2 inset N probN='prob(Norm)'(6.4) /
N 32 cfill=white height=3 position=s;
prob(Norm) 0.0228 Label Hours='Hours';
0 run;
0 0.8 1.6 2.4 3.2 4 4.8
Hours

Baseball Data from 1986-1987 › A Normal Distribution œ

150 250

124 205
125 200
103 164
Number of Players

100
150
Count

75
100 81
50 42 45

50 32
25 N 500
N 322 6 11
8 prob(Norm) 0.8278 1
prob(Norm) 0.0004 0
0 12 24 36 48 60 72 84 96
0 50 100 150 200 250
Random Normal Deviate
Hits

Panel 2:
• normal ... normal Two "normal" keywords are required. The first is for prob(Norm) and the sec-
ond generates a curve.
• normal(noprint color=CX0386BE w=3) secondary options associated with the second "nor-
mal" keyword assign a color and line thickness to the normal curve. NOPRINT "suppresses ta-
bles summarizing the curve" [9,214].
• inset N probN='prob(Norm)'(6.4) The first part of the inset statement defines and formats
the statistics that are displayed. In this instance N and PROBN are requested. PROBN is also
assigned a label and format.
• cfill=white height=3 position=s Statement options control the appearance and position of
the inset . The background color is set to white with CFILL, text HEIGHT is set to 3 (percent), and
POSITION is set to s(outh).

6
SAS Global Forum 2009 Reporting and Information Visualization

HISTOGRAMS FROM THE PLOTHISTO MACRO

While PROC UNIVARIATE produces quality histograms, PLOTHISTO overcomes the following impediments to histo-
gram generation by using PROC GPLOT instead:
1) PLOTHISTO can work with Version 8 software to generate a histogram with endpoints and fre-
quencies.
2) Endpoints can be calculated from the number of bars supplied to the macro as well as the con-
ventional range statement that takes the form of <m TO n BY increment>.
3) Histograms with uneven-width class boundaries can also be generated.
4) The macro can be used to construct marginal histograms associated with scatter plots or histo-
grams coupled with bar charts that represent discrete percent-based frequency distributions.
A complete listing of the PLOTHISTO macro along with subordinate macros can be found in a zip file associated with
paper #NP03 in the NESUG 2008 proceedings. The zip file is also available by request.
PARAMETERS ASSOCIATED WITH THE PLOTHISTO MACRO
In this section, parameters are listed along with examples and associated output. An additional section is devoted to
showing how bar totals are calculated inside the macro. From the header comments in PLOTHISTO:
Parm Name | Description | Default
---------------| ----------------------------------------------- |--------
inds | input data set |
cgmFile | computer-graphics-metafile name (with path) |
xvar | variable plotted on midpoint axis |
xmin | XMIN (where data MIN is calculated) or a value |XMIN
xmax | XMAX (where data MAX is calculated) or a value |XMAX
xdataOffset | For HistoConfig=1,3: |0
| #Units by which to decrease AND increase |
| underlying X-Axis range |
| For HistoConfig=2: |
| Offset in pct assigned in Axis stmt |
HistoConfig | 1=Number of Bars |1
| 2=BY |
| 3=Inner cutpoints(for uneven bars) |
ConfigInfo | If 1, then actual number of bars desired |
| If 2, Interval(BY-value) is supplied |
| If 3, values demarcated by spaces are for |
| inner cutpoints |
Yorigin | Yorigin is used to redraw the horizontal axis |12
pctSize | In Percent (to match annotate) |5
XaxisLbl | X-Axis-Label |
XvalFmt | Display format for X-axis |best.
YaxisLbl | Y-Axis-Label |Frequency
Yby | By value for Y-axis |
ListFreqsYvN | Display frequencies at bar midpoints (YvN) |N
title1 | title1 text including optional move commands |
| Text is enclosed in quotes. |
Key to understanding how PLOTHISTO works is the relationship between the HISTOCONFIG and CONFIGINFO
parameters. If HISTOCONFIG is set to '1' then CONFIGINFO expects a number representing the number of bars.
Defining a histogram by supplying the number of bars is given priority, because the only textbook that provides de-
tailed instructions for histogram construction uses this method [17, p. 37]. To imitate PROC UNIVARIATE,
HISTOCONFIG should be set to '2' so that CONFIGINFO stores the BY value from the range. For the rarely con-
structed histogram with uneven intervals, HISTOCONFIG is set to '3' with CONFIGINFO containing a list of the inner
class intervals. All three methods use XMIN and XMAX as starting points for setting the minimum of the first class
interval and the maximum of the last class interval.
MACRO CALLS:
Example 1: Recreating the UNIVARIATE histogram in Figure 7 with HISTOCONFIG set to '2':
%PlotHisto(inds=histo.meetings, cgmFile=%str(&outpath.\Fig9b.cgm),
xvar=hours, xmin=0, xmax=4.8, xdataOffset=4, HistoConfig=2, ConfigInfo=0.8,
yorigin=12, pctSize=3.75, XaxisLbl=Hours, YaxisLbl=%str(Number of Meetings),
xValFmt=3.1, yby=2, ListFreqsYvN=Y,
title1=%str(move=(+5pct,+0pct) "Meeting Lengths Histogram Via Macro"));

7
SAS Global Forum 2009 Reporting and Information Visualization

The range is reconstructed internally as &XMIN to &XMAX by &CONFIGINFO. The UNVARIATE and macro generated
histograms appear side by side in Figure 9.

Figure 9. UNIVARIATE and macro generated histograms are almost identical when HISTOCONFIG is set to '2'.
Meeting Lengths Histogram via UNIVARIATE Meeting Lengths Histogram Via Macro
10 8 8
8
8 8
8
6
Number of Meetings

Number of Meetings
6
6
6
4 4
4 4 4
4

2 2
2 2

0
0 0.8 1.6 2.4 3.2 4 4.8 0
0.0 0.8 1.6 2.4 3.2 4.0 4.8
Hours Hours

Since PLOTHISTO makes use of the axis statement, both axes labels can be emboldened to set them apart from
their corresponding axis values.
Example 2: Generating n-bar Histograms with HISTOCONFIG set to '1':
%PlotHisto(inds=work.hitsAndRuns, cgmFile=%str(&outpath.\Fig10a.cgm),
xvar=hits, xmin=0, xmax=250, xdataOffset=1,
HistoConfig=1, ConfigInfo=6, yorigin=12, pctSize=3,
XaxisLbl=Hits, YaxisLbl=%str(Frequency), xValFmt=3., yby=25,
ListFreqsYvN=Y,
title1=%str(move=(+10pct,+0pct) "Baseball Data from 1986-1987"));
%PlotHisto(inds=work.hitsAndRuns, cgmFile=%str(&outpath.\Fig10b.cgm),
xvar=hits, xmin=0, xmax=250, xdataOffset=1,
HistoConfig=1, ConfigInfo=9, yorigin=12, pctSize=3,
XaxisLbl=Hits, YaxisLbl=%str(Frequency), xValFmt=3., yby=25,
ListFreqsYvN=Y,
title1=%str(move=(+10pct,+0pct) "Baseball Data from 1986-1987"));
The only changes required for generating the two histograms in Figure 10 are highlighted in the source code above.
Use HISTOCONFIG='1' when the number of bars in a histogram is more important than the intervals that define the
class boundaries.

Figure 10. With HISTOCONFIG set to '1' for an n-bar histogram, CONFIGINFO=6 or 9 produces two histograms with 6 and 9
bars each.
Baseball Data from 1986-1987 Baseball Data from 1986-1987
125 75 73

108

100 56 55
85 52
50 48
74
Frequency

Frequency

75

50
25 21
25 25
25
8 7
5 2
0 0
0 42 83 125 167 208 250 0 28 56 83 111 139 167 194 222 250
Hits Hits

8
SAS Global Forum 2009 Reporting and Information Visualization

Example 3: Generating an uneven interval histogram with HISTOCONFIG set to '3':

%PlotHisto(inds=histo.meetings, cgmFile=%str(&outpath.\Fig11b.cgm),
xvar=hours, xmin=0, xmax=4, xdataOffset=3,
HistoConfig=3, ConfigInfo=%str(0.4 1.2 2.0 2.8 3.6), yorigin=12, pctSize=3.75,
XaxisLbl=Hours, YaxisLbl=%str(Number of Meetings), xValFmt=3.1, yby=2,
ListFreqsYvN=N,
title1=%str(move=(+5pct,+0pct) "Meeting Lengths Histogram Via Macro"));

The only way to convert the midpoint histogram from Figure 1 to the endpoint histogram in Figure 11 is to issue a
macro call with HISTOCONFIG set to 3 for uneven intervals.

Figure 11. With HISTOCONFIG set to '3' for an uneven interval histogram, endpoints are fully defined in XMIN, XMAX and
CONFIGINFO when PLOTHISTO is invoked.

Meeting Lengths Via UNIVARIATE Meeting Lengths Via Macro

30
10

25
8

Number of Meetings
20
Percent

6
15

4
10

5 2

0
0 0.8 1.6 2.4 3.2 4 0
0.0 0.4 1.2 2.0 2.8 3.6 4.0
Hours Hours

HOW PLOTHISTO WORKS:

A task-oriented approach is taken to show how the PLOTHISTO macro works. Key to understanding the macro is a
two-step algorithm that is used for displaying endpoints along the horizontal axis of a histogram. For additional infor-
mation about the algorithm, see Generate a Customized Axis Scale with Uneven Intervals in SAS® Automatically [3].
The algorithm is adapted to work with both even and uneven intervals when histograms are constructed. The exam-
ples in this section are based on the invocation of PLOTHISTO for the uneven axis histogram displayed in Figure 11.
Again, a complete listing of the PLOTHISTO macro along with subordinate macros can be found in a zip file associ-
ated with paper #NP03 in the NESUG 2008 proceedings. The zip file is also available by request.
Task #1: Build the XFMT format from XMIN, XMAX, HISTOCONFIG, and CONFIGINFO
Use macro parameters to calculate cut-points that represent intermediate endpoints sandwiched between XMIN and
XMAX. While processing varies by histogram type (HISTOCONFIG), the code for PROC FORMAT listed below will
always execute.
proc format;
value xfmt &xmin -< &cut1 = "&xmin"
%let ncut_1=%eval(&ncut-1);
%do i= 1 %to &ncut_1;
%let iplus1 = %eval(&i+1);
&&cut&i -< &&cut&iplus1 = "&&cut&i"
%end;
&&cut&ncut - &xmax = "&&cut&ncut"
;
run;

When xmin=0, xmax=4, HistoConfig=3, and ConfigInfo=%str(0.4 1.2 2.0 2.8 3.6) Xfmt is defined as:
Proc format;
value xfmt
0 -< 0.4 = "0" 0.4 -< 1.2 = "0.4" 1.2 -< 2.0 = "1.2"
2.0 -< 2.8 = "2.0" 2.8 -<3.6 = "2.8" 3.6 - 4 = "3.6" ;
run;

XFMT also plays a central role in tasks #3 and #4 below.

9
SAS Global Forum 2009 Reporting and Information Visualization

Task #2:Generate then hide a Conventional Axis with nested macro: MKUNDERLYINGSCALE
When CONFIGINFO is set to 1 (n-bar) or 3 (uneven scale), the macro MKUNDERLINGSCALE is invoked. This
macro makes use of XMIN, XMAX, and XDATAOFFSET sent to PLOTHISTO.
axis2 label=none w=1 value=none major=none minor=none
origin=(,&yorigin.)
%if &histoConfig eq 2 %then
order=(&xmin to &xmax by &ConfigInfo) offset=(&xdataOffset.pct,);
%else
offset=(0pct,)
order=(%MkUnderlyingScale(calcXMin=&xMin, calcXMax=&XMax, OffSet=&xdataOffset));
;
--------------------------------------------------------------------------------------------------------------------------------------------
• label=none major=none minor=none value=none erases the axis completely, leaving only a
single horizontal line. Even though the axis is erased, the ORDER and ORIGIN options remain in
effect. Otherwise the algorithm wouldn't work.
• origin=(,&yorigin)YORIGIN must match the value for YORIGIN in the macro %unevenInter-
valAxis where the displayed X-axis is redrawn via ANNOTATE.
• %MkUnderlyingScale is a macro function that returns an order statement in a range format. For
example for MEETINGS run, that would be -0.16 to 4.16 by 0.04
• &calcXMin, &calcXMax in the case of the MEETINGS data these macro variables resolve to 0.0
and 4.0.
• OffSet=&xdataOffset extends the axis range by +/- 3 (+1) units or 0.16 hours where a unit is
defined as 0.04 in the %getIncr macro contained within MKUNDERLYINGSCALE. The increase
in range is needed to accommodate a text size of 3.75 percent.
Task #3:Generate a Display-Axis with the UNEVENINTERVALAXIS macro
XFMT is used to create a control-out data set that serves as input to the XAXISTICKS data set. Relevant code from
PLOTHISTO:
proc format library=WORK
cntlout=XaxisTicks(keep=start end);
select xfmt;
run;
data XaxisTicks(keep=xtick);
set XaxisTicks;
xtick=input(left(start),best.); output;
xtick=input(left(end),best.); output;
run;
%UnevenIntervalAxis(inDS=xAxisTicks, xvar=xtick, pctSize=&pctSize, xlabel=&XaxisLbl,
yOrigin=&yOrigin, xvalfmt=&xvalfmt.)

A partial listing of the UnevenIntervalAxis macro that uses annotate macros to create tick marks, associated axis val-
ues, and the axis label appears below. Again, full code listings can be found in the zip file.
%macro UnevenIntervalAxis(inDS=, xvar=, pctSize=, xlabel=, yOrigin=, XvalFmt=);
< Local macro variables are assigned and a select distinct in PROC SQL yields DISTINCTXTICK from INDS >
data annoAxisX;
%dclanno;
length text $30;
set distinctXtick end=last;
%system(2,3,3);
%move(xtick, &yOrigin);
%draw(xtick, &yOrigin - &tickLength., black, 1, 0.04);
%label(xtick, &yOrigin - &LabelYpos., DisplayX, black, 0, 0, &pctSize, Hwcgm001,5);
if last then do;
%system(1,3,3);
%label(50, &yOrigin - &axisLabelPos., "&xLabel", black, 0, 0, &pctSize, Hwcgm002,5);
end;
run;
--------------------------------------------------------------------------------------------------------------------------------------------
• %system(2,3,3) is an annotate macro that translates positional parameters to XYS, YSYS and
HSYS coordinate systems. For axis ticks and value labels an XSYS value of '2' uses absolute
values from the data area whereas a value of '3' for YSYS and HSYS translates assigned num-
bers to percentages of the graphics output area. With YSYS set to '3', TICKLENGTH and
LABELYPOS can be accurately subtracted from YORIGIN which is also defined as a percent.
• %system(1,3,3) XSYS, here is changed from '2' to '1' (percent of data area) so that the axis la-
bel is centered on the horizontal axis when the corresponding X coordinate is set to 50.

10
SAS Global Forum 2009 Reporting and Information Visualization

The adjusted axis in Figure 12 highlights the completion of Tasks #2 and #3 above.

Figure 12. The display axis from %UnevenIntervalAxis overlays a grayed-out (usually invisible) axis where the ORDER
option is filled in with an invocation of the %MkUnderlyingScale macro function.

Underlying range = -0.16 TO 4.16 hours BY 0.04 hours

Count of Meetings by their Length

10
9

8 8
8
Number of Meetings

4
4

2
2
1

0 - - - - 000000000000000000000000011111111111111111111111112222222222222222222222222333333333333333333333333344444
0000. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . 000112223344455666778889900011222334445566677888990001122233444556667788899000112223344455666778889900011
1100048260482604826048260482604826048260482604826048260482604826048260482604826048260482604826048260482604826
6284

-0.16 0.0 0.4 1.2 2.0 2.8 3.6 4.0 4.16

Task #4: Create a Plot Data Set by Binning the Input Data with XFMT.
SAS code plus input and output data are listed in this section to demonstrate how the binning uses XFMT to create a
data set amenable to plotting.

/* 1) APPLY XFMT TO THE XVAR COORDINATES */ /* 3) CREATE PLOTDS FROM FREQDS */

data inds(keep=xx &xvar); data pltds(keep=xx yy);
set &inds; set freqDS end=last;
xx=input(put(&xvar,xfmt.),best.); lagy=lag(_freq_);
run; if _n_ = 1 then do;
proc sort data=inds; yy = 0; output;
by &xvar; end;
run; else do;
yy = lagy; output;
/* 2) GET FREQUENCIES FOR RESPONSE AXIS */ yy = 0; output;
proc summary data=inds nway; end;
class xx; if last then do;
output out=freqDS; yy = _freq_; xx = &xmax; output;
run; yy = 0; output;
data freqDS; end;
set freqDS; run;
if _freq_ eq . then _freq_=0;
run;

XX and YY become the plot variables. Zeros are interspersed with actual values for YY when PLTDS is created so
that the symbol statement works as expected when INTERPOLATE= is set to STEPRJ.

From the sorted version of the input data below, it can be seen that there are a lot of tied meeting lengths. Ties
should not be confused with binning. The distinction is addressed in Figure 13.

11
SAS Global Forum 2009 Reporting and Information Visualization

Input Data Sorted: HISTO.MEETINGS Intermediate Output: Final Output: PLTDS

FREQDS
OBS Hours OBS Hours Obs xx _FREQ_ Obs xx yy
------------ ------------ --- --- ------ --- --- --
1 0.0 1 1 0.0 0
1 0.25 17 2.00 2 0.4 9 2 0.4 1
2 0.50 18 2.00 3 1.2 4 3 0.4 0
3 0.50 19 2.50 4 2.0 8 4 1.2 9
4 0.50 20 2.50 5 2.8 8 5 1.2 0
5 0.50 21 2.50 6 3.6 2 6 2.0 4
6 0.50 22 2.50 7 2.0 0
7 0.50 23 3.00 8 2.8 8
8 0.50 24 3.00 9 2.8 0
9 1.00 25 3.00 10 3.6 8
10 1.00 26 3.00 11 3.6 0
11 1.50 27 3.50 12 4.0 2
12 1.50 28 3.50 13 4.0 0
13 1.50 29 3.50
14 1.50 30 3.50
15 2.00 31 4.00
16 2.00 32 4.00

XFMT for the MEETINGS data set is displayed again to show how the binning pictured in Figure 13 works:
Proc format;
value xfmt
0 -< 0.4 = "0" 0.4 -< 1.2 = "0.4" 1.2 -< 2.0 = "1.2"
2.0 -< 2.8 = "2.0" 2.8 -<3.6 = "2.8" 3.6 - 4 = "3.6" ;
run;
Minimum (0) and the maximum (4) are inclusive (>= or <=) whereas intermediate endpoints are exclusive (<). With
this set up, all intermediate points within a member class are set to the value of the left-most endpoint.

Figure 13. XFMT provides the foundation for binning in the PLOTHISTO macro. Using a format means that the generated histo-
gram conforms to the requirement that "each measurement falls into one and only one measurement class" [17,p37].

Count
Raw ofData
Meetings by their Length
with Duplicates ™ Binned Count of Meetings
Data positioned by their Length
at Left-Most Endpoints š
10 10

8 8
Number of Meetings
Number of Meetings

6 6

4 4

2 2

0 0
0.25 0.50 1.00 1.50 2.00 2.50 3.00 3.50 4.0 0.0 0.4 1.2 2.0 2.8 3.6 4.0
Hours Hours

Count of Meetings
Histogram by their
with Binned DataLength › Count ofwith
Histogram Meetings by Binned
Raw and their Length
Data œ
10 10
9 9

8 8 8 8
8 8
Number of Meetings
Number of Meetings

6 6

4 4 4 4 4 4 4
4 4

2 2 2 2
2 2
1 1
1

0
0 0.25 0.5 1 1.5 2 2.5 3 3.5 4
0.0 0.4 1.2 2.0 2.8 3.6 4.0
0.0 0.4 1.2 2.0 2.8 3.6 4.0
Hours Hours

12
SAS Global Forum 2009 Reporting and Information Visualization

The input data are represented by a needle plot in the first panel of Figure 13. A second needle plot shows the calcu-
lated frequencies at the designated intervals, whereas the histogram in panel #3 is generated from the plot data set
containing the interleaved zeros. The fourth panel is a composite of the first three graphs.
Task #5: Color the bars and Generate a Plot
The AREAS= option in the PLOT statement of PROC GPLOT is not satisfactory for coloring the bars of a histogram.
Bar outlines are overwritten! Multiple calls to GREPLAY are convoluted, so the best solution involves the creation of a
second ANNOTATE data set from the plot data set. :
data annoBarFill;
%dclanno;
%system(2,2,3);
set pltds;
xx1=lag(xx); yy1=lag(yy); xx2=xx; yy2=yy;
if xx1 ne . AND yy1 eq 0;
%bar(xx1,yy1,xx2,yy2,graycc,0,solid);
run;
Since annotate macros such as %bar contain an implicit "when=before", the bars are colored before they are outlined.
Now the histogram is ready to be plotted with PROC GPLOT:
proc gplot data=pltds %if &ListFreqsYvN eq N %then anno=annoAxisX; %else anno=annoText; ;
plot yy*xx /vaxis=axis1
haxis=axis2
noframe
anno=annoBarFill;
run;
--------------------------------------------------------------------------------------------------------------------------------------------
• &ListFreqsYvN Not described in this summary is the option for labeling the histogram bars. The
method is very similar to the one used for generating UNIVARIATE histograms.
• anno=annoAxisX %else anno=annoText ANNOTEXT augments ANNOAXISX with a code ex-
tension for the midpoint frequency labels. Both data sets incorporate a call to
UNEVENINTERVALAXIS. The call to MKUNDERLYINGSCALE is embedded in the axis2 state-
ment shown earlier.
• anno=annoBarFill Both the GPLOT and PLOT statements can support the ANNO= option.
Thus the bars can be colored by a separate annotate data set.

AUGMENTED GRAPHS THAT USE THE PLOTHISTO MACRO

MARGINAL HISTOGRAMS FOR DATA OVERLAY
The upper triangular matrix for baseball stats [6] from Multiple-Plot Displays: Simplified with Macros [4] and repro-
duced in Figure 14 provides the motivation for generating marginal histograms as a way to manage data overlay.
Coordinates for individual players remain indistinguishable in the second panel, and league affiliation (+ or X) is
blurred.

Figure 14. An upper triangular matrix is derived from the rectangular matrix that originally appeared in SAS® System for Statisti-
cal Graphics by Michael Friendly.

Upper Triangular Scatterplot Matrix of Baseball Data +

130 ++
352 ++ X XXXX+ XX+X+
+X+X+
+ +X+ XX+ +
++
X + XXXX X + +
XXX+++++
X
X
+ +
X
X+++ X
X+ +
X+
+
X
+
+ X ++ +
XX+ +
X+ X
X++
X X
X
XX+ X
X
+X+
XX+ XX+
+
+ +
+X +X+X ++ +++X+
+ ++X
X
X+
X++XXX+X+
XX
X
X
+
XX
+
+ X+
XX
+XXX+
+ X +
XX
+ X
+
XX
X
+ +
XX+ X+
+
++
+
X
X+ +
X
++X
+
X X +
+X+++ XXXX+X+
++
+
X+
X+
+
XX
+
X
X
X
+ X+
+X+ ++
X
++
X
X+
X
+
X
X +
+
X
+
X
X
X
X+
X
++
X
XXX+
X
X
++
+
XX+
X +X +
XX+
+
X X+
X++X+
X+X
X+
+
X+
++
X X+
X+
X
X+X
+X+
X+
X
X+ X++
X+
X+
X+
X+
+XX X X
++X+
X
+
X +++ +
+X
+
+++
+
+
X+
X
X
+
X
XX
+ +
X
+
X
X
X+ ++X
X
+X+
+ +
X+
++
X X+ X+ ++X+
batavgc +X
+X
++X
XXXX +
X
+X
+
X+
X
++
XX
+X
+
X
+
+
X+
X
X
XX
X
+
X
XX
+
+
X
X
+
X
+
X
X
+
X
+
X
X
+
X
X
+
X
+
++
X
+ +
++ X+ X +X
+
X
+
XX
+
XXX
+
+
+
+
X
XXX
+X+++
+
+X
+
+
+
+
X
+X X + +
X
+
+XX+++++++X +
X
X
+
X
+
+
+
X
+
X
X
+
+
X
+
+
+
++
+
XX
X
X
X
+
X
+
X
+
X
X
+
X
X
X
+
X
+
X
XX
+
X
+
X
+
X
X
+
X
X
X
X
+
X
X
+
++ + X++
X
++
X
+
+
X
X
+
X
X +
X
+X+
X
+ +
X
+
+
+
+
XX
X
++
X
++
X
+
X
+
+
X++X
X
+ ++
+ +
X
+
+
+
X
X+
+
+
X+
X
+
X
X+
X X
X
+
X+
X
+X
+
X+
X+
X+
X
+
+
X+
X
X
+
X+
X+X
X+
X+
+
X
X
X+
XX
X
++
X+
+ X
+
X+
X +
X
+
+
X
X++
X XX+
+
X
X+
+X + +
X
X
+
X
+
X
X
+
+
X
+
+
X
+
X
+
+
X
+
X
X
+X
++
XX
X
++
+
X
+XX
X
X
+
X
+
X
X
+
XX
+
X
+
X
+
+X++X
+X
+
+
X
++
X
X
X
X+
X
X
X
+
X
X
+ X
+
X+
X
+
X
XX
+ +
+XX+X X
X
++ +++
+
XX
+X+
X
X
+++ X +
X
+
XX X
X
X XX
X +
X+
X++ ++
XXX
X + ++X
++ +
++ ++
143 +
+ +
+XX +X++XX+X +X
X
X +
runs
+
X + X+X
+
X+
X
+
+X+
X
X
X
+
X X+
+
+ XX
+
+ ++
+
+
+
+
X++X
X ++X
XX+XXXXX
+ ++XXXX+
130
++X +
X+++
+
X X+
+++X+ +X++
X+
+X + ++ + X ++X+ + XX X
X
+
X+
X
++ ++X+
++ X+
+X
+
X
++
++
XX+XX+ XXX+
+
X+
X X +X +
+
X+
X+++
++
+ XX +
X+
X++
++ +++ XXXX
++ ++ ++
+++
X+
X ++
X+
+ +X + X+X
+
+
+ ++
+XX
+ XX
+
+
X+
X
X++
XX
+
X
runs +X+
XX+
X
+
X+
+X
X
+
X
+
X
X
+
+
X
+
+
X+
++
+
+
+
X
X
X
XXX+
X
+
X
+
X
+
X+
X
X
+XX+
+
X+
X+
++
+
+X+
X
+
X++X
X+
X
X+
X+
+
+
X
X+
XX +
XX+
X+
++
XXX ++ + + +
X X
+
X
+
X
+X
X
+
+
++++X+XX+
X+
+X+
X+
+XXX+
+
X
+XX+
X
+
XX
X
+
+X+X + XX
++++
XX
++XX+X+ X
X +
X+++
+
X
X
+
+
X+
X
+X
X+
X
XX
X
+ X
+
+
+++
X+X
X XXX XX
+ X
+ +X
X++ X+ X XX +
+
XX XX
X+
+ +
+++
X+ XX
X
++X+
+ +
+
X+ XX ++ + X+
X+ +X + XX+ X
X XX
+
X++XX+X
X X X X+ ++ XX XX X +XX
X
X X
X
+
+ +
+++
+X+
+
X
X
+X
X
X
+
X
+
X
+
X
X
X
X
+
X
X+
X
X
X+
X
++
XX
+X++
X
+
X
+
X +
X
+
X
X
X+
X+
X
X
X
X
+++
+
+
X+
X
+X
X
+
X
X
X
+
X+
+
X+
X++
+
X+
X+XX
X
X+
+X+
X X+
X+ X
X+
+
+
X
+
X
X
X
+
X
X
+
X
+
X
X+
X+
X
X+X
X
+
X+
+X+++
X
XXX
X+X
X
X
X +X+
X X
X
X
XX
X
+
+
X
XX
+
+
XX
XX
0
X
+
+
X
X
+
XX
X
X
+
+
X
X
X
+
X
X
X
+++
X
X
X
X
+
X
+
X
X
+XX
+
+X
X
+X
X
+
+
X
X
X+
X+
X
++
+
X
+
X
X+
+
X
X
X+
+
X+X+
+
++
X
XX+X+
XXX+
X
+ +
X
X++
XX XX
X
+
++
X
+
+
X
+
X
+X
X
+
+
X+
X
++
X
+
X
X XXX
X
X
0 X+
+
X X
+X++X
X X
++
+
X
X +X++ +
X+ X X
X
238 +++
XX X++ + X +
+
++
+XX +
XX X ++++X+X + +
+
X+
X
+X++++ ++XXX + ++
X
+ ++
++
X X X+++ +XX
hits X+
+ XX+
X+
X+
X
X++
+
XX+
+X+
+XXX
X+
X
X+
XX+
++ ++
+++ X ++X
XXX+
X +++
X
+
X
X+X+
X X
+X
+
X+ ++
X
+X+XX +X +X +
X++
X X+
X+ X X+
+ X+ ++ ++
+ + ++ + X
+X+ X++ XX
++ +XXX X + X
X
+ X+
+X X+ X
+
++
+X
X+
X
X+
X+
+
X
+X +
XX+X+
X+ + +
X+
X++ + +
+X+
XX++
X X+
+
+
+XX+X
++
X
X+
+X
XX
X X+
X+
+X+
X
+
XX
+
XX+
++
++
XX+
+
X+ +
X+X+
XX
X+
X+X+ X X
+
+
XX
++
XX+
X
XX X
+
X
XX++
X
X X+ X+
+
X X+ + X+ +
+ X
+ ++XXX+X X
X X XX+ X XX
X X X
X+ +
X+
X+XX
+
XX+X+
X+ +
X+
X+XX X+
+ X X+
+
X X+
+XX
X
X+
X+
++X++X+
XX+X+
+ X+ + XX + +
+X +
X
+
X
X
1 X X
X+
X X X+ X X
X
+X++ +
X+ X X

X
238
hits
24 X +
+++X+ +
XXX++ X++X X
years +
++++
X
X
+ +
X
X+ ++
X+
X
+X+ +
X +++
+X+ X + X+X +
+++X++X+++ + XX
XX
X X +
X
XX +++ ++ XXX +
+XX+X+X
XX X + X
X X+
+
+ XX+
X+ X
XX +XX+X
1 +
X
+
X
+
+
+
+
X
X
+X
X
++
X
+
X
X+
+
X
+
+
X
X+
+
X
+
X
+
X ++
+
+X+
+ X
X
++
X
X
+
X
XX+
X
++ + ++
+
X
X + +
+
+
+
X
X
+
+
X
X
X+
X
+X X X

ADAPTED FROM: "SAS System for Statistical Graphics

salary
2460
1
BY: Michael Friendly
AMERICAN LEAGUE: +
NATIONAL LEAGUE: X 68

13
SAS Global Forum 2009 Reporting and Information Visualization

In Figure 15, marginal histograms provide a summary that partially offsets the degree of overlay in the scatter plot.

Figure 15. Marginal histogram totals must be the same, since points are plotted only when information is available for both
RUNS and HITS. Bar heights are comparable between histograms. GREPLAY was used to create the graph.

Baseball Data from 1986-1987

124
103

42 45

8
150

125

10

100

55
Runs

75

85

50

115

25

56

0
0 50 100 150 200 250
Hits

ADD A BAR CHART TO REPRESENT A DISCRETE PERCENT-BASED FREQUENCY DISTRIBUTION

To get a more detailed view of the input data, a short bar chart is added to the base of the histogram along the hori-
zontal axis. In Figure 16, HITS in the second graph are grouped by rounding to the nearest percent. ADJUSTED
HITS are then calculated from the percents with associated frequencies being translated to relative heights for the bar
chart.

Figure 16. The bar chart confirms that the baseball data are not normally distributed. High frequencies are not clustered in the
middle. Spaces between the bars are visible when the graph is enlarged. The maximum frequency is 12.

Baseball Data from 1986-1987 Baseball Data from 1986-1987

150 150
124 124
125 125
Number of Players

103 103
Number of Players

100 100

75 75

50 42 45
50 42 45

25
25 8
N 322
8
prob(Norm) 0.0004 0 12 12
0 0 0
0 50 100 150 200 250
0 50 100 150 200 250
Hits Hits

14
SAS Global Forum 2009 Reporting and Information Visualization

In Figure 17, the bar chart from a normal distribution is more balanced with higher frequencies moving towards the
center of the plot.

Figure 17. The bar chart in the second panel confirms that random normal deviates are, in fact, normally distributed. Now the
maximum frequency in the bar chart is 24.

A Normal Distribution A Normal Distribution

250 250
205 205
200 200
164 164
150 150
Count

Count
100 100 81
81

50 32
50 32
6 11
N 500
11 1
6 prob(Norm) 0.8278 1 0 24 24
0 0 0
12 24 36 48 60 72 84 96
12 24 36 48 60 72 84 96
Random Normal Deviate Random Normal Deviate

SUMMARY AND RECENT EXTENSIONS

Instructions have been provided in the paper for generating textbook histograms in both the UNIVARIATE and
GPLOT procedures. PROC UNIVARIATE is easy to use and can be enhanced with a normal plot overlay. However,
the ENDPOINTS=option in UNIVARIATE is only available in Version 9.13 SAS, and it is not possible to graph n-bar
histograms or histograms with uneven intervals in PROC UNIVARIATE.

While the PLOTHISTO macro that uses PROC GPLOT is more versatile, it is also more complex. However, by re-
viewing the description of how PLOTHISTO works on pages 9-13 in the paper along with a full listing of the source
code in the zip file associated with paper #NP03 in the NESUG 2008 proceedings, it will be possible to generate his-
tograms with ease from PROC GPLOT.

The PLOTHISTO macro has also been recently extended to produce both EMF and CGM output. Normal and
Gamma curves can now be superimposed over GPLOT generated histograms, and subgroup histograms, identical in
concept to subgroup bar charts, are also included in the updated macro. In addition, a separate macro has been writ-
ten to produce a histogram from summary data. Output from the new and recently updated macros will be shown in
the presentation, and a follow-up paper that shows how the new features work is being considered for presentation at
the upcoming NESUG conference.

COPYRIGHT STATEMENT
The paper, Using SAS® Software to Generate Textbook-Style Histograms, along with all associated files in the
NESUG 2008 proceedings is protected by copyright law. This means if you would like to use part or all of the original
ideas or text from these documents in a publication where no monetary profit is to be gained, you are welcome to do
so. All you need to do is to cite the paper in your reference section. For ALL uses that result in corporate or individual
profit, written permission must be obtained from the author. Conditions for usage have been modified from
http://www.whatiscopyright.org.

REFERENCES
[1] Nguyen, Chauthi. Histogram of Numeric Data Distribution from the UNIVARIATE Procedure. Proceedings of the
th
20 Annual Northeast SAS Users Group Conference. Baltimore, MD, 2007, paper #NP12.
[2] Miron, Thomas. The How-To Book for SAS/GRAPH Software. Cary, NC: SAS Institute Inc., 1995.
[3] Watts, Perry. Generate a Customized Axis Scale with Uneven Intervals in SAS® Automatically. Proceedings of
the SAS® Global Forum 2009 Conference. Washington, DC, 2009, paper #192-2009.
[4] Watts, Perry. Multiple-Plot Displays: Simplified with Macros. Cary, NC: SAS Institute Inc., 2002.
WEB CITATIONS:
[5] http://en.wikipedia.org/wiki/Histogram. Histogram: From Wikipedia, the free encyclopedia. The histogram is
defined and compared to a bar chart.

15
SAS Global Forum 2009 Reporting and Information Visualization

[6] http://lib.stat.cmu.edu/datasets/baseball.data. From StatLib --- DataSets Archive. This was the 1988 ASA Graph-
ics Section Poster Session dataset, organized by Lorraine Denby.
SAS INSTITUTE REFERENCES:
[7] SAS Institute Inc. SAS/GRAPH® 9.1 Reference, Volumes 1, 2, and 3, Cary NC: SAS Institute Inc., 2004.
[8] SAS Institute Inc. SAS® Procedures Guide, Version 8, Cary NC: SAS Institute Inc., 1999.
[9] SAS Institute Inc. Base SAS® 9.1.3 Procedures Guide, Volume 3: CORR, FREQ, and UNIVARIATE Procedures,
Cary NC: SAS Institute Inc., 2004.
Statistics Textbooks
[10] Blalock, Hubert M. Social Statistics. New York, NY: McGraw-Hill Book Company, Inc., 1960.
[11] Croxton, Frederick E. and Dudley J. Cowden. Applied General Statistics: Second Edition. New York, NY: Pren-
tice-Hall, Inc., 1955.
[12] Deming, W. Edwards. Making Things Right. Statistics: A Guide to the Unknown. Ed. Judith M. Tanur, et al. San
Francisco, CA: Holden-Day, Inc., 1972. 229-236.
[13] Efron, Bradley Bootstrap Methods: Another Look at the Jackknife. Breakthroughs in Statistics Volume II: Meth-
dology and Distribution. Ed. Samuel Kotz and Norman L. Johnson. New York, NY: Springer-Verlag New York,
Inc., 1992. 569-593.
[14] Freudenthal, Hans. Probability and Statistics. New York, NY: Elsevier Publishing Company, 1965.
[15] Jaeger, Richard M. Statistics A Spectator Sport: Second Edition. Newbury Park, CA: SAGE Publications, Inc.,
1990.
[16] Kvanli, Alan H., C. Stephen Guynes, Robert J. Pavur. Introduction to Business Statistics: A Computer Integrated
Approach. St. Paul, MN: West Publishing Company, 1986.
[17] McClave, James T. and P. George Benson. Statistics for Business and Economics: Third Edition. San Francisco,
CA: Dellen Publishing Company, 1985.
[18] Moore, David S. Statistics Concepts and Controversies: Second Edition. San Francisco, CA: W. H. Freeman and
Company, 1979.
[19] Mosteller, Frederick and David L. Wallace. Deciding Authorship. Statistics: A Guide to the Unknown. Ed. Judith
M. Tanur, et al. San Francisco, CA: Holden-Day, Inc., 1972. 164-175.
[20] Yule, G. Udny and M. G. Kendall. An Introduction to the Theory of Statistics. New York, NY: Hafner Publishing
Company, 1950.

WHAT'S IN THE NESUG 2008 PROCEEDINGS FOR NP03 OR AVAILABLE BY REQUEST:

1) The MEETINGS and BASEBALL data sets
2) SAS Programs
•PlotHisto.sas
Subordinate macros used by PLOTHISTO:
•mkUnderlyingScale.sas
•getIncr.sas
•getAxisMax.sas
•getAxisMin.sas
•UnevenIntervalAxis.sas
The calling program:
•HistoDemo.sas

TRADEMARK CITATION
SAS and all other SAS Institute Inc. product or service names are registered trademarks or trademarks of SAS
Institute Inc. in the USA and other countries. ® indicates USA registration.
Other brand and product names are registered trademarks or trademarks of their respective companies.

CONTACT INFORMATION
The author welcomes feedback via email at perryWatts@comcast.net. A text version of the source code is available
upon request.

16
SAS Global Forum 2009 Reporting and Information Visualization

APPENDIX: TABLES FOR TEXTBOOK LISTINGS BY GRAPH TYPES

Table 1. Horizontal Axis Types: Endpoints vs. Midpoints

Endpoints vs Midpoints
Textbook ID
(Reference #, Title) Type Page Comments

[10] Social Statistics. Endpoints 40-41

Endpoints 42 Juxtaposes a midpoint frequency polygon over
an endpoint histogram
[11] Applied General Statistics: Sec- Endpoints 74 Again, a midpoint frequency polygon is juxta-
ond Edition. posed over an endpoint histogram.
75 The frequency polygon is plotted separately;
however the endpoint axis is left in tact.
[12] Making Things Right. Midpoints 230 Midpoints introduce ambiguity into this graph
which is reproduced in the paper.
[13] Bootstrap Methods: Another Endpoints 589
Look at the Jackknife
[14] Probability and Statistics Endpoints 20-21
[15] Statistics A Spectator Sport: Endpoints 17
Second Edition
[16] Introduction to Business Statis- Endpoints 14-15 While the graphs are plotted with endpoints, the
tics: A Computer Integrated Ap- data is listed with midpoints using MINITAB.
proach.
[17] Statistics for Business and Eco- Endpoints 38,43
nomics: Third Edition
[18] Statistics Concepts and Contro- Midpoints 159
versies: Second Edition. 162 Midpoints introduce ambiguity into this graph
which is reproduced in the paper.
[19] Deciding Authorship. Endpoints 171-
172
[20] An Introduction to the Theory of Midpoints 79 Juxtaposes a midpoint frequency polygon over a
Statistics. midpoint histogram.
Endpoints 90 The endpoint histogram summarizes the number
of deaths from scarlet fever at five year intervals
from ages 5-60. Because of the increased inci-
dence, 0 to 5 years is subdivided into yearly
intervals making this an uneven-interval histo-
gram.

17
SAS Global Forum 2009 Reporting and Information Visualization

Table 2. Vertical Axis Types: Frequencies vs. Percents. (Relative Frequency X 100 = Percent)
Frequencies vs. Percents
Textbook ID
(Reference #, Title) Type Page Comments

[10] Social Statistics. Both 40-42 Percents are listed next to fre-
quencies on the vertical axis (an
effective technique)
[11] Applied General Statistics: Second Frequency 74
Edition.
[12] Making Things Right. Frequency 230
[13] Bootstrap Methods: Another Look at Frequency 589
the Jackknife
[14] Probability and Statistics Frequency 20-21
[15] Statistics A Spectator Sport: Second Frequency 17
Edition
[16] Introduction to Business Statistics: A Frequency 14
Computer Integrated Approach. Relative Frequency 15
[17] Statistics for Business and Econom- Relative Frequency 38
ics: Third Edition Frequency 38
Cumulative Relative 43
Frequency
[18] Statistics Concepts and Controver- Frequency 159
sies: Second Edition. Relative Frequency 162
[19] Deciding Authorship.. Proportion 171-172
[20] An Introduction to the Theory of Frequency 79
Statistics. Frequency 90

18