R Companion - Clewer & Scarisbrick (2001)

R Companion material for Clewer, Alan G., and David H. Scarisbrick. Practical statistics
and experimental design for plant and crop science. John Wiley & Sons, 2001.
Rodrigo R Amadeu
rramadeu at ufl dot edu
September 28th 2018

Contents
Chapter 2: Basic Statistical Calculations 2

Chapter 3: Basic Data Summary 4

Chapter 4: The Normal Distribution, the t-Distribution and Confidence Intervals 12

Chapter 5: Introduction to Hypothesis Testing 24

Chapter 6: Comparison of Two Independent Sample Means 26

Chapter 7: Linear Regression and Correlation 28

Chapter 8: Curve Fitting 35

Chapter 9: The Completely Randomized Design 39

Chapter 10: The Randomized Block Design 57

Chapter 11: The Latin Square Design 67

Chapter 12: Factorial Experiments 79

Chapter 13: Comparison of Treatment Means 89

Chapter 14: Checking the Assumptions and Transformation of Data 97

Chapter 15: Missing Values and Incomplete Blocks 110

Chapter 16: Split Plot Designs 117

R codes to perform most of the statistical analysis of the Clewer & Scarisbrick (2001) from Chapters 1 to 16.
For interpretation and theoretical explanation you should go to the book. If you find any issue, comment,
or error, please send me an e-mail.

1
Chapter 2: Basic Statistical Calculations

Data input in R

Data in R are stored in vectors, to build a vector ‘x’ with the elements from the Example 2.1 in the book
(i.e., 14.8, 15.2, 17.4, 11.6, and 12.5), you can use the function c(), where c stands for combine values into
a vector or list. An example:
x = c(14.8, 15.2, 17.4, 11.6, 12.5)

To visualize the built data, in the RStudio you can look at the Enviroment tab (upper right) which shows
you all the loaded data into your R environment. Another option is to explicitly call it or print it:
x

## [1] 14.8 15.2 17.4 11.6 12.5

print(x)

## [1] 14.8 15.2 17.4 11.6 12.5

Another way to input the data in R is to import from a file. In the RStudio, you can go to Import Dataset
option under the Environment tab and select you data type. Each data type requires a specific command
with specific arguments.

Basic Statistical Calculations

Mean

There are several ways to compute the same value in ‘R’, it follows some of the ways to compute the sample
mean:
You can use R as a calculator and explicitly compute the mean, the signals follow the standard of other
statistical softwares:
(14.8 + 15.2 + 17.4 + 11.6 + 12.5)/5

## [1] 14.3
Or you can use built-in function on your vector x, as:
sum(x)/length(x) #sum all the elements of 'x' and divide such value for its length (n)

## [1] 14.3
mean(x) #compute the mean of x

## [1] 14.3
Hereafter, we show just compact ways to compute different statistics:
median(x) #median of x

## [1] 14.8
var(x) #sample variance of x

## [1] 5.3
var(x)*(length(x)-1) #corrected sum of squares of x

## [1] 21.2

2
sd(x) #sample standard deviation of x

## [1] 2.302173
sd(x)/mean(x) #coefficient of variation of x

## [1] 0.1609911
There is no built-in function to directly compute the coefficient of variation (CV) in R. Loading an R package
is a way to have more functions and dataset to work with. There are thousands of available R packages in
different repositories for different objectives. Here we show how to install and load the ‘sjstats’ packages
which has a built-in function to compute the CV.
install.packages('sjstats') #install sjstats packages
library(sjstats) #loading the package

You can also install a package using RStudio interface, go to the Tools -> Install Packages and use the
search.
Now we have the function cv() available in our R enviroment. To look at its details and to use it:
?cv
cv(x)

## [1] 0.1609911

Weighted mean

x = c(3.5, 4.8, 5.2) #sample means

n = c(4, 5, 10) #sample sizes
?weighted.mean
weighted.mean(x,n)

## [1] 4.736842

3
Chapter 3: Basic Data Summary

3.2: Frequency distributions (discrete data)

Example 3.1: Data input number of tillers

x = c(1, 1, 1, 2, 2, 2, 2,
3, 3, 3, 3, 3, 3, 3,
3, 4, 4, 4, 4, 4, 4,
4, 4, 4, 4, 4, 4, 4,
4, 4, 4, 5, 5, 5, 5,
5, 5, 5, 5, 5, 5, 5,
5, 5, 6, 6, 6, 6, 6,
6, 6, 6, 6, 7, 7, 7,
7, 7, 8, 8)

Basic data summary

sum(x)

## [1] 269
length(x)

## [1] 60
mean(x)

## [1] 4.483333
var(x)

## [1] 2.762429

Frequency table

table(x)

## x
## 1 2 3 4 5 6 7 8
## 3 4 8 16 13 9 5 2

Bar plot

counts <- table(x)

barplot(counts)

4
 15
10
5
0

1 2 3 4 5 6 7 8

# Customizing it a little bit

barplot(counts, xlab = "Number of Fillers (x)", ylab = "Frequency (y)", space = 0.5, ylim=c(0,15))

5
 14
12
10
Frequency (y)

8
6
4
2
0

1 2 3 4 5 6 7 8

Number of Fillers (x)

The mode

sort(counts) #the mode is 4 with 16 counts

## x
## 8 1 2 7 3 6 5 4
## 2 3 4 5 8 9 13 16
median(x)

## [1] 4

3.3 Frequency distributions (continuos data)

Example 3.2: Data input barley yield

x = c(95, 70, 68, 88, 79, 92, 64, 83, 67, 63,
56, 70, 53, 78, 71, 62, 42, 80, 50, 68,
78, 104, 62, 66, 90, 86, 66, 82, 83, 56,
82, 90, 71, 77, 93, 68, 91, 98, 79, 75,
92, 93, 73, 79, 95, 78, 77, 108, 86, 87,
68, 68, 49, 75, 82, 61, 68, 65, 56, 96,
52, 61, 87, 79, 64, 64, 84, 63, 64, 44,
87, 63, 65, 64, 81, 72, 62, 58, 84, 67)

6
length(x) #checking dimension

## [1] 80

The histogram

hist(x)

Histogram of x
25
20
Frequency

15
10
5
0

40 50 60 70 80 90 100 110

# Customizing it a little bit

hist(x, xlab="Yield", main=NULL)

7
 25
20
Frequency

15
10
5
0

40 50 60 70 80 90 100 110

Yield

Quartiles and Ranges

quantile(x,0.25) #Q1

## 25%
## 64
quantile(x,0.75) #Q3

## 75%
## 84
IQR(x) #IQR (i.e., Q3-Q1)

## [1] 20
max(x)

## [1] 108
min(x)

## [1] 42
range(x)

## [1] 42 108

8
Other graphical methods

boxplot(x) #boxplot
90 100
80
70
60
50
40

stripchart(x,method = "stack") #dotplot

9
 40 50 60 70 80 90 100 110

stem(x) #stem and leaf plot

##
## The decimal point is 1 digit(s) to the right of the |
##
## 4 | 249
## 5 | 0236668
## 6 | 1122233344444556677888888
## 7 | 00112355778889999
## 8 | 012223344667778
## 9 | 00122335568
## 10 | 48

Descriptive statistics

The R default summary:

## R Default:
length(x) #N

## [1] 80
summary(x)

## Min. 1st Qu. Median Mean 3rd Qu. Max.

## 42.00 64.00 72.50 73.96 84.00 108.00

10
sd(x) #standard deviation

## [1] 14.02389
var(x) #variance

## [1] 196.6695
mean(x) #mean

## [1] 73.9625
mean(x, trim=0.05) #trimmed mean by 0.05

## [1] 73.97222
There are several packages that provide different types of summary statistics, to see some of them click here.

11
Chapter 4: The Normal Distribution, the t-Distribution and Con-
fidence Intervals

Simulating data in R

To simulate data in R, you need to choose a distribution (probability density function),its parameters, and
how many data points (n) to sample from this distribution.
Here you can find several probability density functions that R can handle. The core idea on simulating data
from such distributions is to know what each one of the simulation functions does (functions r..., p...,
q...,and d...). Knowing this, you can simulate or extract data from different probability density function.
Here we present the four functions of the normal (Gaussian) distribution:
##Distribution Family functions

rnorm

randomly generated n numbers based on a Normal Distribution with mean mean and standard deviation sd:
rnorm(n = 10, mean = 0, sd = 1) #“r”: random, randomly generated numbers f

## [1] -1.06918582 -0.34706905 0.77175552 -0.13966630 0.85168930

## [6] -0.41434788 1.34921727 1.79967071 0.07179162 -0.75991644

pnorm

What is the probability to have values lower than q based on a Normal Distribution with mean mean and
standard deviation sd:
pnorm(q = 0, mean = 0, sd = 1) #“p”: probability, cumulative density function

## [1] 0.5

qnorm (inverse of pnorm)

What is the value (q) correspondent to the probability p to have values lower based on a Normal Distribution
with mean mean and standard deviation sd:
qnorm(p = 0.5, mean = 0, sd = 1) #“q”: quantiles, cumulative density function (quantiles)

## [1] 0

dnorm

What is the density of correspondent to the point x based on a Normal Distribution with mean mean and
standard deviation sd.
dnorm(x = 1, mean = 0, sd = 1)

## [1] 0.2419707

12
Visualizing the distribution

An way to visualize a distribution is to randomly sample a large number of data points from it, then, to
estimate its density estimates, and to plot it in a density plot.
x = rnorm(n = 10000, mean = 0, sd = 1)
dens_x <- density(x)
plot(dens_x)

density.default(x = x)
0.4
0.3
Density

0.2
0.1
0.0

−4 −2 0 2 4

N = 10000 Bandwidth = 0.1431

Example 4.1

x <- rnorm(n = 1000000, mean = 80, sd = 5)

hist(x)

13
 Histogram of x
150000
100000
Frequency

50000
0

60 70 80 90 100

dens_x <- density(x)

plot(dens_x)

14
 density.default(x = x)
0.08
0.06
Density

0.04
0.02
0.00

60 70 80 90 100

N = 1000000 Bandwidth = 0.284

We can estimate some of the properties of the Normal Distribution with simulations:
(2) About 67% of the population is within one standard deviation of the mean:
## How many elements are within 1 sd from the mean (i.e., 75 < x < 85)
interval <- which(x > 75 & x < 85)
length(interval)/length(x)

## [1] 0.682468
(3) 90% of the population is within 1.645 standard deviations of the mean
interval <- which(x > (80 - 1.645*5) & x < 80 + (1.645*5))
length(interval)/length(x)

## [1] 0.900097
(4) 95% of the population is within 1.96 standard deviations of the mean, and hence only 2.5% of the
population have values which are greater than 1.96 standard deviations above the mean.
interval <- which(x > (80 - 1.96*5) & x < 80 + (1.96*5))
length(interval)/length(x)

## [1] 0.950017
interval <- which(x > 80 + (1.96*5))
length(interval)/length(x)

## [1] 0.02498
(5) 99% of the population is within 2.576 standard deviations of the mean

15
interval <- which(x > (80 - 2.576*5) & x < 80 + (2.576*5))
length(interval)/length(x)

## [1] 0.989906
(6) 99.9% of the population is within 3.29 standard deviations of the mean
interval <- which(x > (80 - 3.29*5) & x < 80 + (3.29*5))
length(interval)/length(x)

## [1] 0.999029

Example 4.2

To find the θ(z) in the table of the normal distribuction function in R, you can use:
pnorm(q = -1.60, mean = 0, sd = 1) #find the theta(-1.60) value

## [1] 0.05479929
Otherwise, you can just plug the parameters distribution and avoid the use of the table. Find the proportion
of plants having heights using pnorm function: (1) less than 72 cm
pnorm(q = 72, mean = 80, sd = 5)

## [1] 0.05479929
This value corresponds to the area of the following colored area in probability density plot:
## Warning: Removed 340 rows containing missing values (position_stack).

16
 0.08

0.06
Probability Density

0.04

0.02

0.00

50 55 60 65 70 75 80 85 90 95 100 105 110

Height (cm)

(2) between 82 and 87 cm

pnorm(q = 87, mean = 80, sd = 5) - pnorm(q = 82, mean = 80, sd = 5) #attention for the minus signal

## [1] 0.2638216
This value corresponds to the area of the following colored area in probability density plot:
## Warning: Removed 460 rows containing missing values (position_stack).

17
 0.08

0.06
Probability Density

0.04

0.02

0.00

50 55 60 65 70 75 80 85 90 95 100 105 110

Height (cm)

(3) between 72 and 82 cm

pnorm(q = 82, mean = 80, sd = 5) - pnorm(q = 72, mean = 80, sd = 5) #attention for the minus signal

## [1] 0.6006224
This value corresponds to area of the following colored area in probability density plot:
## Warning: Removed 409 rows containing missing values (position_stack).

18
 0.08

0.06
Probability Density

0.04

0.02

0.00

50 55 60 65 70 75 80 85 90 95 100 105 110

Height (cm)

Example 4.3

(a) Above what height are the top 20% of plants?

qnorm(p = 0.8, mean = 80, sd = 5) #80% of the points are below, then, 20% are above.

## [1] 84.20811
Top 20% of the data is the following colored area in probability density plot:
## Warning: Removed 298 rows containing missing values (position_stack).

19
 0.08

0.06
Probability Density

0.04

0.02

0.00

50 55 60 65 70 75 80 85 90 95 100 105 110

Height (cm)

(b) Below what height are the bottom 4% of plants?

qnorm(p = 0.04, mean = 80, sd = 5) #80% of the points are below, then, 20% are above.

## [1] 71.24657
Bottom 4% of the data is the following colored area in probability density plot:
## Warning: Removed 348 rows containing missing values (position_stack).

20
 0.08

0.06
Probability Density

0.04

0.02

0.00

50 55 60 65 70 75 80 85 90 95 100 105 110

Height (cm)

Normal probability plot

## Generating some data

x <- rnorm(n = 50, mean = 10, sd = 1)
qqnorm(x)

21
 Normal Q−Q Plot
12
11
Sample Quantiles

10
9
8
7

−2 −1 0 1 2

Theoretical Quantiles

4.5 Example

x = c(72.3, 78.9, 82.6, 71.8, 86.1, 80.5, 72.0, 91.8, 77.3, 88.2)
qt(p = 0.975, df = 9) #take the t-table value

## [1] 2.262157
## To find the interval you can carry out a t-test
t.test(x, conf.level = 0.95) #CI 95%

##
## One Sample t-test
##
## data: x
## t = 35.864, df = 9, p-value = 5.042e-11
## alternative hypothesis: true mean is not equal to 0
## 95 percent confidence interval:
## 75.09445 85.20555
## sample estimates:
## mean of x
## 80.15
t.test(x, conf.level = 0.99) #CI 95%

##
## One Sample t-test

22
##
## data: x
## t = 35.864, df = 9, p-value = 5.042e-11
## alternative hypothesis: true mean is not equal to 0
## 99 percent confidence interval:
## 72.88714 87.41286
## sample estimates:
## mean of x
## 80.15

4.6 Example

x = c(171.8, 267.7, 274.7, 203.2, 208.6, 267.2, 184.1, 234.5)

qt(p = 0.975, df = 7) #take the t-table value

## [1] 2.364624
## To find the interval you can carry out a t-test
t.test(x, conf.level = 0.95) #CI 95%

##
## One Sample t-test
##
## data: x
## t = 15.877, df = 7, p-value = 9.536e-07
## alternative hypothesis: true mean is not equal to 0
## 95 percent confidence interval:
## 192.7453 260.2047
## sample estimates:
## mean of x
## 226.475

23
Chapter 5: Introduction to Hypothesis Testing

5.1 Example

x = c(2.6, 2.1, 2.5, 2.4, 1.9, 2.3)

t.test(x, mu = 2.0, conf.level = 0.95) #CI 95%

##
## One Sample t-test
##
## data: x
## t = 2.818, df = 5, p-value = 0.0372
## alternative hypothesis: true mean is not equal to 2
## 95 percent confidence interval:
## 2.026341 2.573659
## sample estimates:
## mean of x
## 2.3
## p-value = 0.0373 which is lower than 5%, therefore, we reject H0 at the 5% level.

5.1 Example

x = c(8.1, 8.7, 9.2, 7.8, 8.4, 9.4)

t.test(x, mu = 8, conf.level = 0.95)

##
## One Sample t-test
##
## data: x
## t = 2.3595, df = 5, p-value = 0.0648
## alternative hypothesis: true mean is not equal to 8
## 95 percent confidence interval:
## 7.94631 9.25369
## sample estimates:
## mean of x
## 8.6
## computed p-value = 0.0648 which is higher than 5%, therefore, we don't reject H0 at the 5% level.

5.2 Example

A <- c(17.8, 18.5, 12.2, 19.7, 10.8, 11.9, 15.6, 12.5)

B <- c(14.7, 15.2, 12.9, 18.3, 10.1, 12.2, 13.5, 9.9)
diff <- A-B #compute the difference
t.test(diff)

##
## One Sample t-test
##
## data: diff
## t = 2.8446, df = 7, p-value = 0.02488

24
## alternative hypothesis: true mean is not equal to 0
## 95 percent confidence interval:
## 0.2573084 2.7926916
## sample estimates:
## mean of x
## 1.525
## computed p-value = 0.02488 which is lower than 5%, therefore, we reject H0 at the 5% level.

25
Chapter 6: Comparison of Two Independent Sample Means

6.1 Example

Control = c(2.65, 3.28, 2.62, 2.84, 2.61, 2.80, 2.23, 2.45, 2.95, 3.12)
Growth = c(3.02, 2.21, 2.29, 3.21, 3.30, 3.13, 2.86, 3.35, 2.72, 3.16)
mean(Control)

## [1] 2.755
mean(Growth)

## [1] 2.925
Changin the two values
Control = c(2.65, 3.28, 2.62, 2.84, 2.61, 2.80, 3.23, 2.45, 2.95, 3.12)
Growth = c(3.02, 2.21, 2.29, 3.21, 3.30, 3.13, 2.86, 2.35, 2.72, 3.16)
mean(Control)

## [1] 2.855
mean(Growth)

## [1] 2.825

6.2 Example

New = c(2.6, 2.1, 2.5, 2.4, 1.9, 2.3)

Standard = c(1.7, 2.1, 2.0, 1.8, 2.3, 1.6, 2.0, 2.1, 2.2, 1.9)
t.test(New, Standard, var.equal = TRUE)

##
## Two Sample t-test
##
## data: New and Standard
## t = 2.7056, df = 14, p-value = 0.01707
## alternative hypothesis: true difference in means is not equal to 0
## 95 percent confidence interval:
## 0.06840468 0.59159532
## sample estimates:
## mean of x mean of y
## 2.30 1.97
## computed p-value = 0.01707 which is lower than 5%, therefore, we reject H0 at the 5% level.

6.3 Example

var.test(New,Standard)

##
## F test to compare two variances
##
## data: New and Standard

26
## F = 1.3878, num df = 5, denom df = 9, p-value = 0.6296
## alternative hypothesis: true ratio of variances is not equal to 1
## 95 percent confidence interval:
## 0.309462 9.271667
## sample estimates:
## ratio of variances
## 1.387755
## computed p-value = 0.6296 which is higer than 5%, therefore, we don't reject H0 at the 5% level.

27
Chapter 7: Linear Regression and Correlation

7.1 Example

Length = c(22, 23, 25, 27, 30, 30, 35, 38, 45,
50, 51, 52, 55, 59, 60, 60, 62, 65,
67, 68, 70, 75, 78, 79, 80, 81,
82, 84, 87, 88, 89, 92, 105, 107)

Area = c(31, 36, 35, 36, 50, 49, 52, 56, 55,
68, 80, 76, 80, 78, 96, 100, 86, 106,
108, 96, 98, 102, 96, 100, 104, 110,
116, 120, 126, 125, 132, 138, 139, 142)

## The Linear Model (regression)

model = lm(Area ~ Length)
summary(model)

##
## Call:
## lm(formula = Area ~ Length)
##
## Residuals:
## Min 1Q Median 3Q Max
## -13.6951 -4.3027 -0.3905 4.0217 14.2925
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 5.74914 3.47886 1.653 0.108
## Length 1.33264 0.05215 25.555 <2e-16 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 7.189 on 32 degrees of freedom
## Multiple R-squared: 0.9533, Adjusted R-squared: 0.9518
## F-statistic: 653.1 on 1 and 32 DF, p-value: < 2.2e-16
## Analysis of Variance
anova(model)

## Analysis of Variance Table

##
## Response: Area
## Df Sum Sq Mean Sq F value Pr(>F)
## Length 1 33750 33750 653.07 < 2.2e-16 ***
## Residuals 32 1654 52
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## Residual Diagnostics
plot(model)

28
 Residuals vs Fitted
15

16 18
10
5
Residuals

0
−5
−15

23

40 60 80 100 120 140

Fitted values
lm(Area ~ Length)

29
 Normal Q−Q

16
2

18
Standardized residuals

1
0
−1
−2

23

−2 −1 0 1 2

Theoretical Quantiles
lm(Area ~ Length)

30
 Scale−Location
16 18 23
1.2
Standardized residuals

0.8
0.4
0.0

40 60 80 100 120 140

Fitted values
lm(Area ~ Length)

31
 Residuals vs Leverage
0.5
2
Standardized residuals

32
1
0
−1

33
−2

Cook's distance 23

0.00 0.02 0.04 0.06 0.08 0.10 0.12

Leverage
lm(Area ~ Length)
hist(model$residuals)

32
 Histogram of model$residuals
10
8
Frequency

6
4
2
0

−15 −10 −5 0 5 10 15

model$residuals

7.2 Example

x = c(0, 25, 50, 75, 100, 125)

y = c(3.70, 4.45, 4.75, 5.20, 5.15, 4.95)
model = lm(y ~ x)
summary(model)

##
## Call:
## lm(formula = y ~ x)
##
## Residuals:
## 1 2 3 4 5 6
## -0.37143 0.12714 0.17571 0.37429 0.07286 -0.37857
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 4.071429 0.249066 16.347 8.2e-05 ***
## x 0.010057 0.003291 3.056 0.0378 *
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 0.3441 on 4 degrees of freedom
## Multiple R-squared: 0.7002, Adjusted R-squared: 0.6252
## F-statistic: 9.341 on 1 and 4 DF, p-value: 0.03779

33
7.3 Example

anova(model)

## Analysis of Variance Table

##
## Response: y
## Df Sum Sq Mean Sq F value Pr(>F)
## x 1 1.10629 1.10629 9.3414 0.03779 *
## Residuals 4 0.47371 0.11843
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
Graphics
plot(x,y)
abline(model)
5.0
4.5
y

4.0

0 20 40 60 80 100 120

34
Chapter 8: Curve Fitting

8.1 Example

y = c(3.70, 4.45, 4.75, 5.20, 5.15, 4.95)

x = c(0, 25, 50, 75, 100, 125)
x2 = x^2
model = lm(y ~ x + x2)
summary(model)

##
## Call:
## lm(formula = y ~ x + x2)
##
## Residuals:
## 1 2 3 4 5 6
## -0.0053571 0.0539286 -0.1171429 0.0814286 -0.0003571 -0.0125000
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 3.705e+00 8.012e-02 46.24 2.23e-05 ***
## x 3.202e-02 3.015e-03 10.62 0.00178 **
## x2 -1.757e-04 2.315e-05 -7.59 0.00474 **
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 0.08841 on 3 degrees of freedom
## Multiple R-squared: 0.9852, Adjusted R-squared: 0.9753
## F-statistic: 99.58 on 2 and 3 DF, p-value: 0.001808
anova(model)

## Analysis of Variance Table

##
## Response: y
## Df Sum Sq Mean Sq F value Pr(>F)
## x 1 1.10629 1.10629 141.551 0.001277 **
## x2 1 0.45027 0.45027 57.612 0.004745 **
## Residuals 3 0.02345 0.00782
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
model$fitted.values

## 1 2 3 4 5 6
## 3.705357 4.396071 4.867143 5.118571 5.150357 4.962500
model$residuals

## 1 2 3 4 5
## -0.0053571429 0.0539285714 -0.1171428571 0.0814285714 -0.0003571429
## 6
## -0.0125000000
plot(x,y)

35
 5.0
4.5
y

4.0

0 20 40 60 80 100 120

8.2 Example

x = c(2.5, 3, 7.2, 7.8, 8.3, 9.8,

10.8, 15.5, 24., 31.5, 40.2, 64.4)

y = c(5.5, 7.9, 9.8, 11, 13.6, 10.9,

12.3, 17.5, 20.5, 25.6, 20.4, 26.8)
lnx = log(x)
lny = log(y)

## Model 1: y = a + bx
model1 = lm(y~x)
summary(model1)

##
## Call:
## lm(formula = y ~ x)
##
## Residuals:
## Min 1Q Median 3Q Max
## -4.2375 -1.9219 -0.8859 2.3061 6.2033
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)

36
## (Intercept) 8.9048 1.3901 6.406 7.78e-05 ***
## x 0.3331 0.0539 6.179 0.000104 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 3.306 on 10 degrees of freedom
## Multiple R-squared: 0.7925, Adjusted R-squared: 0.7717
## F-statistic: 38.18 on 1 and 10 DF, p-value: 0.0001042
model1$fitted.values

## 1 2 3 4 5 6 7
## 9.737508 9.904046 11.302967 11.502813 11.669351 12.168966 12.502042
## 8 9 10 11 12
## 14.067502 16.898651 19.396724 22.294489 30.354939
model1$residuals

## 1 2 3 4 5 6
## -4.2375081 -2.0040464 -1.5029673 -0.5028132 1.9306486 -1.2689660
## 7 8 9 10 11 12
## -0.2020424 3.4324984 3.6013488 6.2032756 -1.8944893 -3.5549387
## Model 2: ln(y) = a + ln(x)
model2 = lm(lny~lnx)

8.3 Example

x = c(100, 130, 150, 198, 204, 214, 220, 229, 234,

241, 246, 252, 257, 266, 273, 281, 287, 290, 292)

y = c(0.2, 0.4, 0.8, 1.9, 2.2, 3.2, 3.9, 6.5, 8.5,

10.1, 12.2, 12, 12.4, 13.2, 14.5, 13.3, 13.1, 13, 12.8)

plot(x,y)

37
 10 12 14
8
y

6
4
2
0

100 150 200 250

## Logistic growth curve

model = nls(y~alpha/(1+exp(h+c*x)), start=list(alpha=13.54, h=19.70, c=-0.0864))

38
Chapter 9: The Completely Randomized Design

9.1 Example

From Chapter 9 and beyond we should have our data settled in a data frame (spreadsheet) which is a
combination of vectors of same size. You can create it internally in R or in a externally with the aid of a
spreadsheet software (MS Excel, LibreOffice Calc etc) and import in R with Import Dataset from RStudio.
Below, we build the data frame of Example 9.1. You can read the data in different ways. Here, we read the
data row after row based on Table 9.3.
data = data.frame(variety = c("A", "D", "B", "D", "C",
"C", "D", "D", "A", "D",
"A", "B", "C", "C", "B",
"A", "B", "A", "C", "B"),
yield = c(22.2, 23.9, 24.1, 21.7, 25.9,
18.4, 24.8, 28.2, 17.3, 26.4,
21.2, 30.3, 23.2, 21.9, 27.4,
25.2, 26.4, 16.1, 22.6, 34.8))

## To access the vectors use $ sign

print(data)

## variety yield
## 1 A 22.2
## 2 D 23.9
## 3 B 24.1
## 4 D 21.7
## 5 C 25.9
## 6 C 18.4
## 7 D 24.8
## 8 D 28.2
## 9 A 17.3
## 10 D 26.4
## 11 A 21.2
## 12 B 30.3
## 13 C 23.2
## 14 C 21.9
## 15 B 27.4
## 16 A 25.2
## 17 B 26.4
## 18 A 16.1
## 19 C 22.6
## 20 B 34.8
data$variety

## [1] A D B D C C D D A D A B C C B A B A C B
## Levels: A B C D
data$yield

## [1] 22.2 23.9 24.1 21.7 25.9 18.4 24.8 28.2 17.3 26.4 21.2 30.3 23.2 21.9
## [15] 27.4 25.2 26.4 16.1 22.6 34.8

39
## Always check the structure of your data:
str(data)

## 'data.frame': 20 obs. of 2 variables:

## $ variety: Factor w/ 4 levels "A","B","C","D": 1 4 2 4 3 3 4 4 1 4 ...
## $ yield : num 22.2 23.9 24.1 21.7 25.9 18.4 24.8 28.2 17.3 26.4 ...
#num = quantitative variable
#factor = qualitative variable

Be carefull, always check your data structure, if the variable types are correct, if the number of factor levels
are correct etc.
tapply(data$yield, data$variety, length) #N

## A B C D
## 5 5 5 5
tapply(data$yield, data$variety, mean) #Mean

## A B C D
## 20.4 28.6 22.4 25.0
tapply(data$yield, data$variety, sd) #StDev

## A B C D
## 3.708773 4.118859 2.700926 2.466779
model = aov(yield ~ variety, data = data)
anova(model)

## Analysis of Variance Table

##
## Response: yield
## Df Sum Sq Mean Sq F value Pr(>F)
## variety 3 188.2 62.733 5.6901 0.00756 **
## Residuals 16 176.4 11.025
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
boxplot(yield ~ variety, data = data)

40
 35
30
25
20

A B C D

plot(model)

41
 Residuals vs Fitted

20
6

16
4
Residuals

2
0
−2
−4

22 24 26 28

Fitted values
aov(yield ~ variety)

42
 Normal Q−Q

20
2
Standardized residuals

16
1
0
−1

−2 −1 0 1 2

Theoretical Quantiles
aov(yield ~ variety)

43
 Scale−Location
1.5

20

16
3
Standardized residuals

1.0
0.5
0.0

22 24 26 28

Fitted values
aov(yield ~ variety)

44
 Constant Leverage:
Residuals vs Factor Levels

20
2
Standardized residuals

16
1
0
−1

variety :
A B C D

Factor Level Combinations

9.2 Example

data = data.frame(variety = c("A","A","A",

"B","B","B","B",
"C","C","C","C",
"D","D","D","D","D"),
yield = c(17.3, 21.2, 16.1,
24.1, 30.3, 26.4, 34.8,
25.9, 18.4, 21.9, 22.6,
23.9, 21.7, 24.8, 28.2, 26.4))

tapply(data$yield, data$variety, length) #N

## A B C D
## 3 4 4 5
tapply(data$yield, data$variety, mean) #Mean

## A B C D
## 18.2 28.9 22.2 25.0
tapply(data$yield, data$variety, sd) #StDev

## A B C D
## 2.666458 4.692547 3.075711 2.466779

45
model = aov(yield ~ variety, data = data)
anova(model)

## Analysis of Variance Table

##
## Response: yield
## Df Sum Sq Mean Sq F value Pr(>F)
## variety 3 214.92 71.640 6.4638 0.007499 **
## Residuals 12 133.00 11.083
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
boxplot(yield ~ variety, data = data)
35
30
25
20

A B C D

plot(model)

46
 Residuals vs Fitted
6

7
4
Residuals

2
0
−2
−4

9
4
−6

18 20 22 24 26 28

Fitted values
aov(yield ~ variety)

47
 Normal Q−Q

7
2
Standardized residuals

1
0
−1

9
4

−2 −1 0 1 2

Theoretical Quantiles
aov(yield ~ variety)

48
 Scale−Location
7
4
1.2
Standardized residuals

9
0.8
0.4
0.0

18 20 22 24 26 28

Fitted values
aov(yield ~ variety)

49
 Residuals vs Leverage

7
2

0.5
Standardized residuals

2
1
0
−1

Cook's distance 4
−2

0.5

0.00 0.05 0.10 0.15 0.20 0.25 0.30

Leverage
aov(yield ~ variety)

agricolae package

There are several R package for Experimental Design Analysis for Agricultural and Plant Breeding experi-
ments (more here). agricolae is by far the most-used package from this task view (status: October 2017).
Here, we show how to do the RCD analysis using agricolae
To install and load agricolae
#install.packages("agricolae")
library(agricolae)

Sampling the treatments

variety = c("A","B","C","D")

field_design = design.crd(variety, r=5)

field_design$book

## plots r variety
## 1 101 1 D
## 2 102 2 D
## 3 103 1 C
## 4 104 1 B
## 5 105 1 A

50
## 6 106 2 A
## 7 107 3 D
## 8 108 2 C
## 9 109 3 A
## 10 110 4 D
## 11 111 3 C
## 12 112 4 A
## 13 113 4 C
## 14 114 2 B
## 15 115 5 A
## 16 116 3 B
## 17 117 5 D
## 18 118 4 B
## 19 119 5 B
## 20 120 5 C

Data analysis

data = data.frame(variety = c("A", "D", "B", "D", "C",

"C", "D", "D", "A", "D",
"A", "B", "C", "C", "B",
"A", "B", "A", "C", "B"),
yield = c(22.2, 23.9, 24.1, 21.7, 25.9,
18.4, 24.8, 28.2, 17.3, 26.4,
21.2, 30.3, 23.2, 21.9, 27.4,
25.2, 26.4, 16.1, 22.6, 34.8))

model = aov(yield ~ variety, data = data)

anova(model)

## Analysis of Variance Table

##
## Response: yield
## Df Sum Sq Mean Sq F value Pr(>F)
## variety 3 188.2 62.733 5.6901 0.00756 **
## Residuals 16 176.4 11.025
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## Coefficient of Variation
cv.model(model)

## [1] 13.77756
## Least Significant Difference Analysis
LSD = LSD.test(model, "variety", console = TRUE)

##
## Study: model ~ "variety"
##
## LSD t Test for yield
##
## Mean Square Error: 11.025
##
## variety, means and individual ( 95 %) CI

51
##
## yield std r LCL UCL Min Max
## A 20.4 3.708773 5 17.2521 23.5479 16.1 25.2
## B 28.6 4.118859 5 25.4521 31.7479 24.1 34.8
## C 22.4 2.700926 5 19.2521 25.5479 18.4 25.9
## D 25.0 2.466779 5 21.8521 28.1479 21.7 28.2
##
## Alpha: 0.05 ; DF Error: 16
## Critical Value of t: 2.119905
##
## least Significant Difference: 4.451801
##
## Treatments with the same letter are not significantly different.
##
## yield groups
## B 28.6 a
## D 25.0 ab
## C 22.4 bc
## A 20.4 c
bar.group(LSD$group)

ab
25

bc
c
20
15
10
5
0

B D C A

bar.err(LSD$means)

52
 −−−

−−− −−−
25

−−− −−−
−−− −−−
20

−−−
15
10
5
0

A B C D

plot(LSD, variation="range",las=1)

53
 Groups and Range

40
a
35

30 ab
bc c
25

20

15

B D C A

plot(LSD, horiz=TRUE, variation="SD",las=1)

54
 Groups and Standard deviation

A c

C bc

D ab

B a

15 20 25 30 35

## Tukey's (HSD) Test

HSD.test(model, "variety", console=TRUE)

##
## Study: model ~ "variety"
##
## HSD Test for yield
##
## Mean Square Error: 11.025
##
## variety, means
##
## yield std r Min Max
## A 20.4 3.708773 5 16.1 25.2
## B 28.6 4.118859 5 24.1 34.8
## C 22.4 2.700926 5 18.4 25.9
## D 25.0 2.466779 5 21.7 28.2
##
## Alpha: 0.05 ; DF Error: 16
## Critical Value of Studentized Range: 4.046093
##
## Minimun Significant Difference: 6.008142
##
## Treatments with the same letter are not significantly different.
##
## yield groups

55
## B 28.6 a
## D 25.0 ab
## C 22.4 b
## A 20.4 b

56
Chapter 10: The Randomized Block Design

Randomization

4 treatments and 3 blocks (repetitions)

library(agricolae)
variety = c("A","B","C","D")

field_design = design.rcbd(variety,r = 3)
field_design$book

## plots block variety

## 1 101 1 A
## 2 102 1 D
## 3 103 1 C
## 4 104 1 B
## 5 201 2 D
## 6 202 2 C
## 7 203 2 A
## 8 204 2 B
## 9 301 3 C
## 10 302 3 D
## 11 303 3 B
## 12 304 3 A

10.1 Example

Below, we build the data frame of Example 10.1. Here we are reading the data by columns.
example10.1 = data.frame(variety = c("V1", "V1", "V1",
"V2", "V2", "V2",
"V3", "V3", "V3",
"V4", "V4", "V4"),
block = c("B1", "B2", "B3",
"B1", "B2", "B3",
"B1", "B2", "B3",
"B1", "B2", "B3"),
yield = c(7.4, 6.5, 5.6,
9.8, 6.8, 6.2,
7.3, 6.1, 6.4,
9.5, 8.0, 7.4))
## Visualing the data frame
print(example10.1)

## variety block yield

## 1 V1 B1 7.4
## 2 V1 B2 6.5
## 3 V1 B3 5.6
## 4 V2 B1 9.8
## 5 V2 B2 6.8
## 6 V2 B3 6.2
## 7 V3 B1 7.3
## 8 V3 B2 6.1

57
## 9 V3 B3 6.4
## 10 V4 B1 9.5
## 11 V4 B2 8.0
## 12 V4 B3 7.4
Checking the data frame structure, sums, block and variety totals
str(example10.1)

## 'data.frame': 12 obs. of 3 variables:

## $ variety: Factor w/ 4 levels "V1","V2","V3",..: 1 1 1 2 2 2 3 3 3 4 ...
## $ block : Factor w/ 3 levels "B1","B2","B3": 1 2 3 1 2 3 1 2 3 1 ...
## $ yield : num 7.4 6.5 5.6 9.8 6.8 6.2 7.3 6.1 6.4 9.5 ...
#apply the function sum in yield by block
tapply(example10.1$yield, example10.1$block, sum)

## B1 B2 B3
## 34.0 27.4 25.6
#apply the function mean in yield by block
tapply(example10.1$yield, example10.1$block, mean)

## B1 B2 B3
## 8.50 6.85 6.40
#apply the function sum in yield by variety
tapply(example10.1$yield, example10.1$variety, sum)

## V1 V2 V3 V4
## 19.5 22.8 19.8 24.9
#apply the function mean in yield by variety
tapply(example10.1$yield, example10.1$variety, mean)

## V1 V2 V3 V4
## 6.5 7.6 6.6 8.3

Descriptive plots

boxplot( example10.1$yield ~ example10.1$block)

58
 9
8
7
6

B1 B2 B3

boxplot( example10.1$yield ~ example10.1$variety)

59
 9
8
7
6

V1 V2 V3 V4

10.2 Analysis ignoring blocks

model10.2 = aov(yield ~ variety, data = example10.1)

anova(model10.2)

## Analysis of Variance Table

##
## Response: yield
## Df Sum Sq Mean Sq F value Pr(>F)
## variety 3 6.63 2.2100 1.4516 0.2987
## Residuals 8 12.18 1.5225

10.3 The analysis including blocks

model10.3 = aov(yield ~ block + variety, data = example10.1)

anova(model10.3)

## Analysis of Variance Table

##
## Response: yield
## Df Sum Sq Mean Sq F value Pr(>F)
## block 2 9.78 4.89 12.225 0.007651 **
## variety 3 6.63 2.21 5.525 0.036730 *
## Residuals 6 2.40 0.40

60
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

Residuals and Fitted values

model10.3$residuals

## 1 2 3 4 5 6 7 8 9 10 11 12
## -0.35 0.40 -0.05 0.95 -0.40 -0.55 -0.55 -0.10 0.65 -0.05 0.10 -0.05
model10.3$fitted.values

## 1 2 3 4 5 6 7 8 9 10 11 12
## 7.75 6.10 5.65 8.85 7.20 6.75 7.85 6.20 5.75 9.55 7.90 7.45

Diagnostics Plots

plot(model10.3)

Residuals vs Fitted
1.0

9
0.5
Residuals

0.0
−0.5

6 7 8 9

Fitted values
aov(yield ~ block + variety)

61
 Normal Q−Q

4
2
Standardized residuals

9
1
0
−1

−1.5 −1.0 −0.5 0.0 0.5 1.0 1.5

Theoretical Quantiles
aov(yield ~ block + variety)

62
 Scale−Location
1.5

4
Standardized residuals

9
7
1.0
0.5
0.0

6 7 8 9

Fitted values
aov(yield ~ block + variety)

63
 Constant Leverage:
Residuals vs Factor Levels

4
2
Standardized residuals

9
1
0
−1

block :
B1 B2 B3

Factor Level Combinations

hist(model10.3$residuals,breaks = 10)

64
 Histogram of model10.3$residuals
4
3
Frequency

2
1
0

−0.5 0.0 0.5 1.0

model10.3$residuals

10.8 Comparison of treatment means

10.8.1 LSD Analysis and Confidence Intervals

With agricolae
library(agricolae)
LSD.test(model10.3, "variety", console=TRUE)

##
## Study: model10.3 ~ "variety"
##
## LSD t Test for yield
##
## Mean Square Error: 0.4
##
## variety, means and individual ( 95 %) CI
##
## yield std r LCL UCL Min Max
## V1 6.5 0.9000000 3 5.606514 7.393486 5.6 7.4
## V2 7.6 1.9287302 3 6.706514 8.493486 6.2 9.8
## V3 6.6 0.6244998 3 5.706514 7.493486 6.1 7.3
## V4 8.3 1.0816654 3 7.406514 9.193486 7.4 9.5
##
## Alpha: 0.05 ; DF Error: 6

65
## Critical Value of t: 2.446912
##
## least Significant Difference: 1.26358
##
## Treatments with the same letter are not significantly different.
##
## yield groups
## V4 8.3 a
## V2 7.6 ab
## V3 6.6 b
## V1 6.5 b

Standard Errors and Confidence Intervals (by “hand””)

## Extracting RMS
anova(model10.3)

## Analysis of Variance Table

##
## Response: yield
## Df Sum Sq Mean Sq F value Pr(>F)
## block 2 9.78 4.89 12.225 0.007651 **
## variety 3 6.63 2.21 5.525 0.036730 *
## Residuals 6 2.40 0.40
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## The RMS is the Residual Mean Square error,
## In our ANOVA table, RMS is located in the 3rd row and 3rd column
RMS = anova(model10.3)[3,3]
print(RMS)

## [1] 0.4
r = 3 #number of reps (blocks)

SEM = sqrt(RMS/r)
print(SEM)

## [1] 0.3651484
lower.t = qt(0.025, 6)
upper.t = qt(0.975, 6)

variety.mean = tapply(example10.1$yield, example10.1$variety, mean)

lower.CI = variety.mean + lower.t*SEM

upper.CI = variety.mean + upper.t*SEM

CI.variety = data.frame(mean = variety.mean,

lower = lower.CI,
upper = upper.CI)

66
Chapter 11: The Latin Square Design

Randomization

variety = c("A","B","C","D")

field_design = design.lsd(variety)
field_design$book

## plots row col variety

## 1 101 1 1 B
## 2 102 1 2 A
## 3 103 1 3 C
## 4 104 1 4 D
## 5 201 2 1 C
## 6 202 2 2 B
## 7 203 2 3 D
## 8 204 2 4 A
## 9 301 3 1 D
## 10 302 3 2 C
## 11 303 3 3 A
## 12 304 3 4 B
## 13 401 4 1 A
## 14 402 4 2 D
## 15 403 4 3 B
## 16 404 4 4 C

11.1 Example

Below, we build the data frame of Example 11.1. Here we are reading the data by columns.
example11.1 = data.frame(
row = c("1","1","1","1",
"2","2","2","2",
"3","3","3","3",
"4","4","4","4"),
column = c("1","2","3","4",
"1","2","3","4",
"1","2","3","4",
"1","2","3","4"),
variety = c("C","D","B","A",
"B","A","C","D",
"D","C","A","B",
"A","B","D","C"),
yield = c(16.6,13.9,18.0,21.9,
12.6,17.4,17.9,16.5,
9.3,16.2,18.4,15.5,
15.7,11.3,10.5,13.1))

## Visualing the data frame

print(example11.1)

## row column variety yield

67
## 1 1 1 C 16.6
## 2 1 2 D 13.9
## 3 1 3 B 18.0
## 4 1 4 A 21.9
## 5 2 1 B 12.6
## 6 2 2 A 17.4
## 7 2 3 C 17.9
## 8 2 4 D 16.5
## 9 3 1 D 9.3
## 10 3 2 C 16.2
## 11 3 3 A 18.4
## 12 3 4 B 15.5
## 13 4 1 A 15.7
## 14 4 2 B 11.3
## 15 4 3 D 10.5
## 16 4 4 C 13.1
#apply the function sum in yield by row
tapply(example11.1$yield, example11.1$row, sum)

## 1 2 3 4
## 70.4 64.4 59.4 50.6
#apply the function sum in yield by col
tapply(example11.1$yield, example11.1$col, sum)

## 1 2 3 4
## 54.2 58.8 64.8 67.0
#apply the function sum in yield by variety
tapply(example11.1$yield, example11.1$variety, sum)

## A B C D
## 73.4 57.4 63.8 50.2
#apply the function mean in yield by variety
tapply(example11.1$yield, example11.1$variety, mean)

## A B C D
## 18.35 14.35 15.95 12.55

Model

model11.1 = aov(yield ~ row + column + variety, data = example11.1)

Residual analysis

plot(model11.1)

68
 Residuals vs Fitted
2

13 8
1
Residuals

0
−1

16
−2

10 12 14 16 18 20 22

Fitted values
aov(yield ~ row + column + variety)

69
 Normal Q−Q
2

8 13
Standardized residuals

1
0
−1

16

−2 −1 0 1 2

Theoretical Quantiles
aov(yield ~ row + column + variety)

70
 Scale−Location
13 8
16
1.2
Standardized residuals

0.8
0.4
0.0

10 12 14 16 18 20 22

Fitted values
aov(yield ~ row + column + variety)

71
 Constant Leverage:
Residuals vs Factor Levels
2

8 13
Standardized residuals

1
0
−1

16
−2

row :
1 2 3 4

Factor Level Combinations

Analysis of Variance

anova(model11.1)

## Analysis of Variance Table

##
## Response: yield
## Df Sum Sq Mean Sq F value Pr(>F)
## row 3 52.62 17.5400 7.4217 0.01918 *
## column 3 25.34 8.4467 3.5740 0.08629 .
## variety 3 72.76 24.2533 10.2623 0.00889 **
## Residuals 6 14.18 2.3633
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

Comparision between the treatments

LSD.test(model11.1, "variety", console=TRUE)

##
## Study: model11.1 ~ "variety"
##
## LSD t Test for yield
##

72
## Mean Square Error: 2.363333
##
## variety, means and individual ( 95 %) CI
##
## yield std r LCL UCL Min Max
## A 18.35 2.615977 4 16.46916 20.23084 15.7 21.9
## B 14.35 3.000556 4 12.46916 16.23084 11.3 18.0
## C 15.95 2.033880 4 14.06916 17.83084 13.1 17.9
## D 12.55 3.275668 4 10.66916 14.43084 9.3 16.5
##
## Alpha: 0.05 ; DF Error: 6
## Critical Value of t: 2.446912
##
## least Significant Difference: 2.659903
##
## Treatments with the same letter are not significantly different.
##
## yield groups
## A 18.35 a
## C 15.95 ab
## B 14.35 bc
## D 12.55 c

11.1 Exercise

exercise11.1 = data.frame(
row = c("1","1","1","1","1",
"2","2","2","2","2",
"3","3","3","3","3",
"4","4","4","4","4",
"5","5","5","5","5"),
column = c("1","2","3","4","5",
"1","2","3","4","5",
"1","2","3","4","5",
"1","2","3","4","5",
"1","2","3","4","5"),
variety = c("P2","P3","P5","P1","P4",
"P4","P5","P1","P2","P3",
"P3","P1","P2","P4","P5",
"P5","P2","P4","P3","P1",
"P1","P4","P3","P5","P2"),
yield = c(62.3,61.3,62.5,63.8,75.0,
64.1,68.4,62.9,66.2,77.4,
69.2,55.8,67.8,71.3,74.8,
65.0,68.7,69.8,76.0,70.9,
63.3,75.0,69.3,78.0,75.4))

Model

modelexercise11.1 = aov(yield ~ row + column + variety, data = exercise11.1)

73
Residual analysis

plot(modelexercise11.1)

Residuals vs Fitted
4

11
2
Residuals

0
−2

9
−4

12

60 65 70 75

Fitted values
aov(yield ~ row + column + variety)

74
 Normal Q−Q
1.5

11
Standardized residuals

0.5
−0.5
−1.5

9
12

−2 −1 0 1 2

Theoretical Quantiles
aov(yield ~ row + column + variety)

75
 Scale−Location
12
9
1.2

11
Standardized residuals

0.8
0.4
0.0

60 65 70 75

Fitted values
aov(yield ~ row + column + variety)

76
 Constant Leverage:
Residuals vs Factor Levels

11
Standardized residuals

1
0
−1

9
12
−2

row :
1 2 3 4 5

Factor Level Combinations

Analysis of Variance

anova(modelexercise11.1)

## Analysis of Variance Table

##
## Response: yield
## Df Sum Sq Mean Sq F value Pr(>F)
## row 4 147.81 36.953 3.5607 0.038911 *
## column 4 350.23 87.558 8.4369 0.001768 **
## variety 4 196.74 49.185 4.7393 0.015857 *
## Residuals 12 124.54 10.378
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

Comparision between the treatments

LSD.test(modelexercise11.1, "variety", console=TRUE)

##
## Study: modelexercise11.1 ~ "variety"
##
## LSD t Test for yield
##

77
## Mean Square Error: 10.37793
##
## variety, means and individual ( 95 %) CI
##
## yield std r LCL UCL Min Max
## P1 63.34 5.348177 5 60.20101 66.47899 55.8 70.9
## P2 68.08 4.769382 5 64.94101 71.21899 62.3 75.4
## P3 70.64 6.432962 5 67.50101 73.77899 61.3 77.4
## P4 71.04 4.503665 5 67.90101 74.17899 64.1 75.0
## P5 69.74 6.529012 5 66.60101 72.87899 62.5 78.0
##
## Alpha: 0.05 ; DF Error: 12
## Critical Value of t: 2.178813
##
## least Significant Difference: 4.439207
##
## Treatments with the same letter are not significantly different.
##
## yield groups
## P4 71.04 a
## P3 70.64 a
## P5 69.74 a
## P2 68.08 a
## P1 63.34 b

78
Chapter 12: Factorial Experiments

Randomization

Let’s assume that we would like to use the following factors: Factor 1 with 2 levels Factor 2 with 3 levels
We can use the same previous functions with the 6 possible treatments as follows
factor1 = c("A1","B1")
factor2 = c("A2","B2","C2")
treatments = paste(rep(factor1,each=3),rep(factor2,2),sep="_")

## Randomization in CRD with 3 reps

field_design = design.crd(treatments,r = 3)

## Randomization in RCBD with 3 blocks

field_design = design.rcbd(treatments,r = 3)

## Randomization in Latin Square 6x6

field_design = design.lsd(treatments)

Example 12.3

Hereafter, we will use the function rep to build the data frames
example12.3 = data.frame(
block = rep(c("B1","B2","B3"),each=15),
P = rep(rep(c("P1","P2","P3"),each=5),3),
N = rep(c("N1","N2","N3","N4","N5"),9),
yield = c(0.9,1.2,1.3,1.8,1.1,0.9,1.1,1.3,1.6,1.9,0.9,1.4,1.3,1.4,1.2,
0.9,1.3,1.5,1.9,1.4,0.8,0.9,1.5,1.3,1.6,1.0,1.2,1.4,1.5,1.1,
1.0,1.2,1.4,2.1,1.2,0.8,0.9,1.1,1.1,1.5,0.7,1.0,1.4,1.4,1.3)
)

## Visualing the data frame

print(example12.3)

## block P N yield
## 1 B1 P1 N1 0.9
## 2 B1 P1 N2 1.2
## 3 B1 P1 N3 1.3
## 4 B1 P1 N4 1.8
## 5 B1 P1 N5 1.1
## 6 B1 P2 N1 0.9
## 7 B1 P2 N2 1.1
## 8 B1 P2 N3 1.3
## 9 B1 P2 N4 1.6
## 10 B1 P2 N5 1.9
## 11 B1 P3 N1 0.9
## 12 B1 P3 N2 1.4
## 13 B1 P3 N3 1.3
## 14 B1 P3 N4 1.4
## 15 B1 P3 N5 1.2
## 16 B2 P1 N1 0.9

79
## 17 B2 P1 N2 1.3
## 18 B2 P1 N3 1.5
## 19 B2 P1 N4 1.9
## 20 B2 P1 N5 1.4
## 21 B2 P2 N1 0.8
## 22 B2 P2 N2 0.9
## 23 B2 P2 N3 1.5
## 24 B2 P2 N4 1.3
## 25 B2 P2 N5 1.6
## 26 B2 P3 N1 1.0
## 27 B2 P3 N2 1.2
## 28 B2 P3 N3 1.4
## 29 B2 P3 N4 1.5
## 30 B2 P3 N5 1.1
## 31 B3 P1 N1 1.0
## 32 B3 P1 N2 1.2
## 33 B3 P1 N3 1.4
## 34 B3 P1 N4 2.1
## 35 B3 P1 N5 1.2
## 36 B3 P2 N1 0.8
## 37 B3 P2 N2 0.9
## 38 B3 P2 N3 1.1
## 39 B3 P2 N4 1.1
## 40 B3 P2 N5 1.5
## 41 B3 P3 N1 0.7
## 42 B3 P3 N2 1.0
## 43 B3 P3 N3 1.4
## 44 B3 P3 N4 1.4
## 45 B3 P3 N5 1.3
#apply the function sum in yield by block
tapply(example12.3$yield, example12.3$block, sum)

## B1 B2 B3
## 19.3 19.3 18.1
#apply the function sum in yield by P
tapply(example12.3$yield, example12.3$P, sum)

## P1 P2 P3
## 20.2 18.3 18.2
#apply the function sum in yield by N
tapply(example12.3$yield, example12.3$N, sum)

## N1 N2 N3 N4 N5
## 7.9 10.2 12.2 14.1 12.3

Model

model12.3 <- aov(yield ~ block + P + N + P*N, data=example12.3)

80
Redisual analysis

plot(model12.3)

Residuals vs Fitted

9
34
0.2

10
0.1
Residuals

0.0
−0.1
−0.2

0.8 1.0 1.2 1.4 1.6 1.8 2.0

Fitted values
aov(yield ~ block + P + N + P * N)

81
 Normal Q−Q

9
34
2

10
Standardized residuals

1
0
−1

−2 −1 0 1 2

Theoretical Quantiles
aov(yield ~ block + P + N + P * N)

82
 Scale−Location
1.5

9
34
10
Standardized residuals

1.0
0.5
0.0

0.8 1.0 1.2 1.4 1.6 1.8 2.0

Fitted values
aov(yield ~ block + P + N + P * N)

83
 Constant Leverage:
Residuals vs Factor Levels

9
34
2

10
Standardized residuals

1
0
−1
−2

block :
B1 B2 B3

Factor Level Combinations

Analysis of Variance

anova(model12.3)

## Analysis of Variance Table

##
## Response: yield
## Df Sum Sq Mean Sq F value Pr(>F)
## block 2 0.06400 0.03200 1.6311 0.21377
## P 2 0.16933 0.08467 4.3155 0.02324 *
## N 4 2.49022 0.62256 31.7322 4.946e-10 ***
## P:N 8 1.01511 0.12689 6.4676 8.979e-05 ***
## Residuals 28 0.54933 0.01962
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

Comparision between the treatments

With factorial experiment we need to define for which factor we want to do the comparision
For P
LSD.test(model12.3, trt = "P", console=TRUE)

##

84
## Study: model12.3 ~ "P"
##
## LSD t Test for yield
##
## Mean Square Error: 0.01961905
##
## P, means and individual ( 95 %) CI
##
## yield std r LCL UCL Min Max
## P1 1.346667 0.3542934 15 1.272585 1.420748 0.9 2.1
## P2 1.220000 0.3405877 15 1.145919 1.294081 0.8 1.9
## P3 1.213333 0.2294922 15 1.139252 1.287415 0.7 1.5
##
## Alpha: 0.05 ; DF Error: 28
## Critical Value of t: 2.048407
##
## least Significant Difference: 0.104767
##
## Treatments with the same letter are not significantly different.
##
## yield groups
## P1 1.346667 a
## P2 1.220000 b
## P3 1.213333 b
For N
LSD.test(model12.3, trt = "N", console=TRUE)

##
## Study: model12.3 ~ "N"
##
## LSD t Test for yield
##
## Mean Square Error: 0.01961905
##
## N, means and individual ( 95 %) CI
##
## yield std r LCL UCL Min Max
## N1 0.8777778 0.09718253 9 0.782139 0.9734165 0.7 1.0
## N2 1.1333333 0.17320508 9 1.037695 1.2289721 0.9 1.4
## N3 1.3555556 0.12360331 9 1.259917 1.4511943 1.1 1.5
## N4 1.5666667 0.31622777 9 1.471028 1.6623054 1.1 2.1
## N5 1.3666667 0.26457513 9 1.271028 1.4623054 1.1 1.9
##
## Alpha: 0.05 ; DF Error: 28
## Critical Value of t: 2.048407
##
## least Significant Difference: 0.1352536
##
## Treatments with the same letter are not significantly different.
##
## yield groups
## N4 1.5666667 a
## N5 1.3666667 b

85
## N3 1.3555556 b
## N2 1.1333333 c
## N1 0.8777778 d
For P*N interaction, the agricolae syntax is c("P","N")
LSD.test(model12.3, trt = c("P","N"), console=TRUE)

##
## Study: model12.3 ~ c("P", "N")
##
## LSD t Test for yield
##
## Mean Square Error: 0.01961905
##
## P:N, means and individual ( 95 %) CI
##
## yield std r LCL UCL Min Max
## P1:N1 0.9333333 0.05773503 3 0.7676821 1.0989845 0.9 1.0
## P1:N2 1.2333333 0.05773503 3 1.0676821 1.3989845 1.2 1.3
## P1:N3 1.4000000 0.10000000 3 1.2343488 1.5656512 1.3 1.5
## P1:N4 1.9333333 0.15275252 3 1.7676821 2.0989845 1.8 2.1
## P1:N5 1.2333333 0.15275252 3 1.0676821 1.3989845 1.1 1.4
## P2:N1 0.8333333 0.05773503 3 0.6676821 0.9989845 0.8 0.9
## P2:N2 0.9666667 0.11547005 3 0.8010155 1.1323179 0.9 1.1
## P2:N3 1.3000000 0.20000000 3 1.1343488 1.4656512 1.1 1.5
## P2:N4 1.3333333 0.25166115 3 1.1676821 1.4989845 1.1 1.6
## P2:N5 1.6666667 0.20816660 3 1.5010155 1.8323179 1.5 1.9
## P3:N1 0.8666667 0.15275252 3 0.7010155 1.0323179 0.7 1.0
## P3:N2 1.2000000 0.20000000 3 1.0343488 1.3656512 1.0 1.4
## P3:N3 1.3666667 0.05773503 3 1.2010155 1.5323179 1.3 1.4
## P3:N4 1.4333333 0.05773503 3 1.2676821 1.5989845 1.4 1.5
## P3:N5 1.2000000 0.10000000 3 1.0343488 1.3656512 1.1 1.3
##
## Alpha: 0.05 ; DF Error: 28
## Critical Value of t: 2.048407
##
## least Significant Difference: 0.2342662
##
## Treatments with the same letter are not significantly different.
##
## yield groups
## P1:N4 1.9333333 a
## P2:N5 1.6666667 b
## P3:N4 1.4333333 bc
## P1:N3 1.4000000 c
## P3:N3 1.3666667 c
## P2:N4 1.3333333 c
## P2:N3 1.3000000 c
## P1:N2 1.2333333 c
## P1:N5 1.2333333 c
## P3:N2 1.2000000 cd
## P3:N5 1.2000000 cd
## P2:N2 0.9666667 de
## P1:N1 0.9333333 e

86
## P3:N1 0.8666667 e
## P2:N1 0.8333333 e

Interaction Plots

interaction.plot(example12.3$N,
example12.3$P,
example12.3$yield,
ylab= "Mean Yield",
xlab = "Nitrogen Level",
trace.label = NULL)
1.8

P2
P1
1.6

P3
Mean Yield

1.4
1.2
1.0
0.8

N1 N2 N3 N4 N5

Nitrogen Level

interaction.plot(example12.3$P,
example12.3$N,
example12.3$yield,
ylab= "Mean Yield",
xlab = "Plant Regulator Level",
trace.label = NULL)

87
 1.8

N4
N3
1.6

N2
Mean Yield

N5
1.4

N1
1.2
1.0
0.8

P1 P2 P3

Plant Regulator Level

88
Chapter 13: Comparison of Treatment Means
Since the book does not provide the Example 13.1 original data, here we show how to perfom the comparison
of treatment means using the example 9.1
example9.1 = data.frame(variety = c("A", "D", "B", "D", "C",
"C", "D", "D", "A", "D",
"A", "B", "C", "C", "B",
"A", "B", "A", "C", "B"),
yield = c(22.2, 23.9, 24.1, 21.7, 25.9,
18.4, 24.8, 28.2, 17.3, 26.4,
21.2, 30.3, 23.2, 21.9, 27.4,
25.2, 26.4, 16.1, 22.6, 34.8))
model9.1 = aov(yield~variety,data=example9.1)
anova(model9.1)

## Analysis of Variance Table

##
## Response: yield
## Df Sum Sq Mean Sq F value Pr(>F)
## variety 3 188.2 62.733 5.6901 0.00756 **
## Residuals 16 176.4 11.025
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

13.2.1 Least Significant Difference

library(agricolae)
LSD.test(model9.1, "variety", console=TRUE)

##
## Study: model9.1 ~ "variety"
##
## LSD t Test for yield
##
## Mean Square Error: 11.025
##
## variety, means and individual ( 95 %) CI
##
## yield std r LCL UCL Min Max
## A 20.4 3.708773 5 17.2521 23.5479 16.1 25.2
## B 28.6 4.118859 5 25.4521 31.7479 24.1 34.8
## C 22.4 2.700926 5 19.2521 25.5479 18.4 25.9
## D 25.0 2.466779 5 21.8521 28.1479 21.7 28.2
##
## Alpha: 0.05 ; DF Error: 16
## Critical Value of t: 2.119905
##
## least Significant Difference: 4.451801
##
## Treatments with the same letter are not significantly different.
##
## yield groups

89
## B 28.6 a
## D 25.0 ab
## C 22.4 bc
## A 20.4 c

13.2.2 Tukey’s Honestly Significant Difference Test (HSD) a.k.a. Tukey’s Stu-
dentised Range Test

library(agricolae)
HSD.test(model9.1, "variety", console=TRUE)

##
## Study: model9.1 ~ "variety"
##
## HSD Test for yield
##
## Mean Square Error: 11.025
##
## variety, means
##
## yield std r Min Max
## A 20.4 3.708773 5 16.1 25.2
## B 28.6 4.118859 5 24.1 34.8
## C 22.4 2.700926 5 18.4 25.9
## D 25.0 2.466779 5 21.7 28.2
##
## Alpha: 0.05 ; DF Error: 16
## Critical Value of Studentized Range: 4.046093
##
## Minimun Significant Difference: 6.008142
##
## Treatments with the same letter are not significantly different.
##
## yield groups
## B 28.6 a
## D 25.0 ab
## C 22.4 b
## A 20.4 b

13.2.3 Student-Newman-Keuls Test (SNK)

library(agricolae)
SNK.test(model9.1, "variety", console=TRUE)

##
## Study: model9.1 ~ "variety"
##
## Student Newman Keuls Test
## for yield
##
## Mean Square Error: 11.025
##

90
## variety, means
##
## yield std r Min Max
## A 20.4 3.708773 5 16.1 25.2
## B 28.6 4.118859 5 24.1 34.8
## C 22.4 2.700926 5 18.4 25.9
## D 25.0 2.466779 5 21.7 28.2
##
## Alpha: 0.05 ; DF Error: 16
##
## Critical Range
## 2 3 4
## 4.451801 5.418695 6.008142
##
## Means with the same letter are not significantly different.
##
## yield groups
## B 28.6 a
## D 25.0 ab
## C 22.4 b
## A 20.4 b

13.2.4 Duncan’s Multiple Range Test (DMRT)

library(agricolae)
duncan.test(model9.1, "variety", console=TRUE)

##
## Study: model9.1 ~ "variety"
##
## Duncan's new multiple range test
## for yield
##
## Mean Square Error: 11.025
##
## variety, means
##
## yield std r Min Max
## A 20.4 3.708773 5 16.1 25.2
## B 28.6 4.118859 5 24.1 34.8
## C 22.4 2.700926 5 18.4 25.9
## D 25.0 2.466779 5 21.7 28.2
##
## Alpha: 0.05 ; DF Error: 16
##
## Critical Range
## 2 3 4
## 4.451801 4.668308 4.803648
##
## Means with the same letter are not significantly different.
##
## yield groups
## B 28.6 a

91
## D 25.0 ab
## C 22.4 b
## A 20.4 b

13.2.5 Waller-Duncan’s Bayes MSD test

library(agricolae)
waller.test(model9.1, "variety", console=TRUE)

##
## Study: model9.1 ~ "variety"
##
## Waller-Duncan K-ratio t Test for yield
##
## This test minimizes the Bayes risk under additive loss and certain other assumptions
## ......
## K ratio 100.000000
## Error Degrees of Freedom 16.000000
## Error Mean Square 11.025000
## F value 5.690098
## Critical Value of Waller 2.153000
##
## variety, means
##
## yield std r Min Max
## A 20.4 3.708773 5 16.1 25.2
## B 28.6 4.118859 5 24.1 34.8
## C 22.4 2.700926 5 18.4 25.9
## D 25.0 2.466779 5 21.7 28.2
##
## Minimum Significant Difference 4.5213
## Treatments with the same letter are not significantly different.
##
## yield groups
## B 28.6 a
## D 25.0 ab
## C 22.4 bc
## A 20.4 c

13.4.1 Testing for a Linear Trend in Treatment Means

Example 13.3

example13.3 = data.frame(block=rep(c("A","B","C","D"),each=6),
N=rep(c(0,25,50,75,100,125),4),
yield=c(3.6,4.8,4.4,5.3,4.8,5.0,
4.1,5.1,5.2,5.9,5.5,5.4,
3.2,4.0,4.6,4.6,5.2,4.8,
3.9,3.9,4.8,5.0,5.1,4.6))

Analysis of Variance with N as factor (qualitative)

92
example13.3$N = as.factor(example13.3$N)
model13.3 = aov(yield~block+N,data=example13.3)
anova(model13.3)

## Analysis of Variance Table

##
## Response: yield
## Df Sum Sq Mean Sq F value Pr(>F)
## block 3 2.19 0.730 9.359 0.000988 ***
## N 5 6.32 1.264 16.205 1.397e-05 ***
## Residuals 15 1.17 0.078
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
Analysis of Variance with N as numeric for testing a linear trend. It is necessary to include the treatment at
factor (factor(N)) to count for the treatment deviations in the ANOVA table after N in the formula, i.e., N
+ factor(N). Attention! The opposite factor(N) + N does not result in the same result, here the order
of the summation matters.
example13.3 = data.frame(block=rep(c("A","B","C","D"),each=6),
N=rep(c(0,25,50,75,100,125),4),
yield=c(3.6,4.8,4.4,5.3,4.8,5.0,
4.1,5.1,5.2,5.9,5.5,5.4,
3.2,4.0,4.6,4.6,5.2,4.8,
3.9,3.9,4.8,5.0,5.1,4.6))
model13.3linear = aov(yield~block+N+factor(N),data=example13.3)
anova(model13.3linear)

## Analysis of Variance Table

##
## Response: yield
## Df Sum Sq Mean Sq F value Pr(>F)
## block 3 2.1900 0.7300 9.3590 0.000988 ***
## N 1 4.4251 4.4251 56.7326 1.794e-06 ***
## factor(N) 4 1.8949 0.4737 6.0733 0.004115 **
## Residuals 15 1.1700 0.0780
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

13.4.2 Testing for a Quadratic Trend in Treatment Means

Analysis of Variance with N as numeric for testing a linear and quadratic trends. Note that here we need to
use the I() function to avoid the direct sum of the terms before going into the function. It is necessary to
include the treatment at factor (factor(N)) to count for the treatment deviations in the ANOVA table.
example13.3 = data.frame(block=rep(c("A","B","C","D"),each=6),
N=rep(c(0,25,50,75,100,125),4),
yield=c(3.6,4.8,4.4,5.3,4.8,5.0,
4.1,5.1,5.2,5.9,5.5,5.4,
3.2,4.0,4.6,4.6,5.2,4.8,
3.9,3.9,4.8,5.0,5.1,4.6))

model13.3quadratic = aov(yield~block+I(N)+I(N^2)+factor(N),data=example13.3)
anova(model13.3quadratic)

93
## Analysis of Variance Table
##
## Response: yield
## Df Sum Sq Mean Sq F value Pr(>F)
## block 3 2.1900 0.7300 9.3590 0.0009880 ***
## I(N) 1 4.4251 4.4251 56.7326 1.794e-06 ***
## I(N^2) 1 1.8011 1.8011 23.0907 0.0002315 ***
## factor(N) 3 0.0938 0.0313 0.4008 0.7544699
## Residuals 15 1.1700 0.0780
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
summary(model13.3quadratic)

## Df Sum Sq Mean Sq F value Pr(>F)

## block 3 2.190 0.730 9.359 0.000988 ***
## I(N) 1 4.425 4.425 56.733 1.79e-06 ***
## I(N^2) 1 1.801 1.801 23.091 0.000231 ***
## factor(N) 3 0.094 0.031 0.401 0.754470
## Residuals 15 1.170 0.078
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

13.4.4 Regression in a Two-Factor Experiment

Now we need to have the factors as numeric. The following example test linear trends for each factor and
also for the interaction.
example12.3 = data.frame(
block = rep(c("B1","B2","B3"),each=15),
P = rep(rep(c(1,2,3),each=5),3),
N = rep(c(1,2,3,4,5),9),
yield = c(0.9,1.2,1.3,1.8,1.1,0.9,1.1,1.3,1.6,1.9,0.9,1.4,1.3,1.4,1.2,
0.9,1.3,1.5,1.9,1.4,0.8,0.9,1.5,1.3,1.6,1.0,1.2,1.4,1.5,1.1,
1.0,1.2,1.4,2.1,1.2,0.8,0.9,1.1,1.1,1.5,0.7,1.0,1.4,1.4,1.3)
)

model12.3 = aov(yield~block+P+factor(P)+N+factor(N)+P*N+factor(P*N),data=example12.3)
summary(model12.3)

## Df Sum Sq Mean Sq F value Pr(>F)

## block 2 0.0640 0.0320 1.688 0.2026
## P 1 0.1333 0.1333 7.032 0.0128 *
## factor(P) 1 0.0360 0.0360 1.899 0.1788
## N 1 1.7921 1.7921 94.512 1.25e-10 ***
## factor(N) 3 0.6981 0.2327 12.272 2.33e-05 ***
## factor(P * N) 7 1.0146 0.1449 7.644 3.08e-05 ***
## Residuals 29 0.5499 0.0190
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

94
13.5 Treatment with Structure (Contrasts)

Example 13.5

example13.5 = data.frame(
block = rep(c("B1","B2","B3","B4"),each=5),
herbicide = rep(c("A","B","C","D","E"),4),
yield = c(64,68,71,75,69,
66,64,69,74,71,
68,60,75,78,65,
59,61,68,72,60)
)

model13.5 = aov(yield~block+herbicide,data=example13.5)
anova(model13.5)

## Analysis of Variance Table

##
## Response: yield
## Df Sum Sq Mean Sq F value Pr(>F)
## block 3 99.75 33.25 3.9118 0.0368135 *
## herbicide 4 370.80 92.70 10.9059 0.0005752 ***
## Residuals 12 102.00 8.50
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
Building the contrasts table (as table 13.14)
c1 <- c(4,-1,-1,-1,-1)
c2 <- c(0,3,-1,-1,-1)
c3 <- c(0,0,2,-1,-1)
c4 <- c(0,0,0,1,-1)

## Combine in a matrix
contrasts <- cbind(c1,c2,c3,c4)

## Define contrasts in the object herbicides

contrasts(example13.5$herbicide) <- contrasts

## ANOVA w/ contrasts
model13.5 = aov(yield~block+herbicide,data=example13.5)
summary.aov(model13.5,split =list(herbicide=list(1,2,3,4)))

## Df Sum Sq Mean Sq F value Pr(>F)

## block 3 99.8 33.25 3.912 0.036814 *
## herbicide 4 370.8 92.70 10.906 0.000575 ***
## herbicide: C1 1 64.8 64.80 7.624 0.017245 *
## herbicide: C2 1 161.3 161.33 18.980 0.000934 ***
## herbicide: C3 1 0.2 0.17 0.020 0.890961
## herbicide: C4 1 144.5 144.50 17.000 0.001413 **
## Residuals 12 102.0 8.50
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

95
## If you want to put the contrast names:
summary.aov(model13.5,split =list(herbicide=list("A vs B,C,D,E"=1, "B vs C,D,E" = 2, "C vs D,E"=3, "D vs

## Df Sum Sq Mean Sq F value Pr(>F)

## block 3 99.8 33.25 3.912 0.036814 *
## herbicide 4 370.8 92.70 10.906 0.000575 ***
## herbicide: A vs B,C,D,E 1 64.8 64.80 7.624 0.017245 *
## herbicide: B vs C,D,E 1 161.3 161.33 18.980 0.000934 ***
## herbicide: C vs D,E 1 0.2 0.17 0.020 0.890961
## herbicide: D vs E 1 144.5 144.50 17.000 0.001413 **
## Residuals 12 102.0 8.50
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
Including orthogonal contrasts

96
Chapter 14: Checking the Assumptions and Transformation of
Data
Example 14.1

example14.1 = data.frame(
herbicide = rep(c("A","B","C","D"),6),
yield = c(4,8,25,33,
5,11,28,21,
2,9,20,48,
5,12,15,18,
4,7,14,53,
1,7,30,31)
)

model14.1 = aov(yield~herbicide,data=example14.1)
anova(model14.1)

## Analysis of Variance Table

##
## Response: yield
## Df Sum Sq Mean Sq F value Pr(>F)
## herbicide 3 3361.1 1120.38 17.876 7.022e-06 ***
## Residuals 20 1253.5 62.67
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
plot(model14.1)

97
 Residuals vs Fitted
20

20

12
10
Residuals

0
−10

16
−20

5 10 15 20 25 30 35

Fitted values
aov(yield ~ herbicide)

98
 Normal Q−Q
3

20
Standardized residuals

12
1
0
−1
−2

16

−2 −1 0 1 2

Theoretical Quantiles
aov(yield ~ herbicide)

99
 Scale−Location
20
1.5

16
12
Standardized residuals

1.0
0.5
0.0

5 10 15 20 25 30 35

Fitted values
aov(yield ~ herbicide)

100
 Constant Leverage:
Residuals vs Factor Levels
3

20
Standardized residuals

12
1
0
−1
−2

16

herbicide :
A B C D

Factor Level Combinations

Log transformation
model14.1log = aov(log(yield)~herbicide,data=example14.1)
anova(model14.1log)

## Analysis of Variance Table

##
## Response: log(yield)
## Df Sum Sq Mean Sq F value Pr(>F)
## herbicide 3 19.328 6.4428 34.216 4.54e-08 ***
## Residuals 20 3.766 0.1883
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
plot(model14.1log)

101
 Residuals vs Fitted
0.5

20
0.0
Residuals

−0.5

16
−1.0

21

1.5 2.0 2.5 3.0 3.5

Fitted values
aov(log(yield) ~ herbicide)

102
 Normal Q−Q

20
1
Standardized residuals

0
−1

16
−2

21

−2 −1 0 1 2

Theoretical Quantiles
aov(log(yield) ~ herbicide)

103
 Scale−Location
21
1.5
Standardized residuals

16
20
1.0
0.5
0.0

1.5 2.0 2.5 3.0 3.5

Fitted values
aov(log(yield) ~ herbicide)

104
 Constant Leverage:
Residuals vs Factor Levels

20
1
Standardized residuals

0
−1

16
−2

21
−3

herbicide :
A B C D

Factor Level Combinations

Square-Root transformation
model14.1sqrt = aov(sqrt(yield)~herbicide,data=example14.1)
anova(model14.1sqrt)

## Analysis of Variance Table

##
## Response: sqrt(yield)
## Df Sum Sq Mean Sq F value Pr(>F)
## herbicide 3 54.132 18.0440 30.289 1.235e-07 ***
## Residuals 20 11.914 0.5957
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
plot(model14.1sqrt)

105
 Residuals vs Fitted
1.5

20
12
0.5
Residuals

−0.5
−1.5

16

2 3 4 5

Fitted values
aov(sqrt(yield) ~ herbicide)

106
 Normal Q−Q

20
2

12
Standardized residuals

1
0
−1
−2

16

−2 −1 0 1 2

Theoretical Quantiles
aov(sqrt(yield) ~ herbicide)

107
 Scale−Location
1.5

20
16
12
Standardized residuals

1.0
0.5
0.0

2 3 4 5

Fitted values
aov(sqrt(yield) ~ herbicide)

108
 Constant Leverage:
Residuals vs Factor Levels

20
2

12
Standardized residuals

1
0
−1
−2

16

herbicide :
A B C D

Factor Level Combinations

109
Chapter 15: Missing Values and Incomplete Blocks

Example 15.1

Missing values in R are storage as NA

example15.1 = data.frame(
treatment = rep(c("1","2","3","4"),each=3),
Y1 = c(6.4,NA,5.6,
9.8,8.7,7.2
,7.3,6.1,6.4,
NA,8.0,9.4),
Y2 = c(12.2,13.4,NA,
15.5,16.3,17.8,
10.4,NA,10.6,
NA,16.8,17.8)
)

model15.1 = aov(Y1~treatment,data=example15.1)
anova(model15.1)

## Analysis of Variance Table

##
## Response: Y1
## Df Sum Sq Mean Sq F value Pr(>F)
## treatment 3 13.2223 4.4074 4.8198 0.0487 *
## Residuals 6 5.4867 0.9144
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

Example 15.2

The R default is to use the Adjusted SS when missing values are present.
example15.2 = data.frame(variety = c("V1", "V1", "V1",
"V2", "V2", "V2",
"V3", "V3", "V3",
"V4", "V4", "V4"),
block = c("B1", "B2", "B3",
"B1", "B2", "B3",
"B1", "B2", "B3",
"B1", "B2", "B3"),
yield = c(NA, 6.5, 5.6,
9.8, 6.8, 6.2,
7.3, 6.1, NA,
9.5, 8.0, 7.4))

model15.2 = aov(yield~block+variety,data=example15.2)
anova(model15.2)

## Analysis of Variance Table

##
## Response: yield
## Df Sum Sq Mean Sq F value Pr(>F)

110
## block 2 10.5993 5.2997 18.8792 0.009176 **
## variety 3 6.2938 2.0979 7.4736 0.040742 *
## Residuals 4 1.1229 0.2807
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
Pairwise comparison
library(agricolae)
LSD.test(model15.2,"variety",console=TRUE)

##
## Study: model15.2 ~ "variety"
##
## LSD t Test for yield
##
## Mean Square Error: 0.2807143
##
## variety, means and individual ( 95 %) CI
##
## yield std r LCL UCL Min Max
## V1 6.05 0.6363961 2 5.009825 7.090175 5.6 6.5
## V2 7.60 1.9287302 3 6.750701 8.449299 6.2 9.8
## V3 6.70 0.8485281 2 5.659825 7.740175 6.1 7.3
## V4 8.30 1.0816654 3 7.450701 9.149299 7.4 9.5
##
## Alpha: 0.05 ; DF Error: 4
## Critical Value of t: 2.776445
##
## Groups according to probability of means differences and alpha level( 0.05 )
##
## Treatments with the same letter are not significantly different.
##
## yield groups
## V4 8.30 a
## V2 7.60 ab
## V3 6.70 bc
## V1 6.05 c
HSD.test(model15.2,"variety",console=TRUE)

##
## Study: model15.2 ~ "variety"
##
## HSD Test for yield
##
## Mean Square Error: 0.2807143
##
## variety, means
##
## yield std r Min Max
## V1 6.05 0.6363961 2 5.6 6.5
## V2 7.60 1.9287302 3 6.2 9.8
## V3 6.70 0.8485281 2 6.1 7.3
## V4 8.30 1.0816654 3 7.4 9.5
##

111
## Alpha: 0.05 ; DF Error: 4
## Critical Value of Studentized Range: 5.757058
##
## Groups according to probability of means differences and alpha level( 0.05 )
##
## Treatments with the same letter are not significantly different.
##
## yield groups
## V4 8.30 a
## V2 7.60 ab
## V3 6.70 ab
## V1 6.05 b

15.5 Incomplete Block Designs

15.5.1 Balanced Incomplete Blocks

Randomization
library(agricolae)
treatments = c(1:9)
field.design = design.bib(t=treatments,k=3,r=12)

##
## Parameters BIB
## ==============
## Lambda : 3
## treatmeans : 9
## Block size : 3
## Blocks : 36
## Replication: 12
##
## Efficiency factor 0.75
##
## <<< Book >>>
field.design

## $parameters
## $parameters$design
## [1] "bib"
##
## $parameters$trt
## [1] 1 2 3 4 5 6 7 8 9
##
## $parameters$k
## [1] 3
##
## $parameters$serie
## [1] 2
##
## $parameters$seed
## [1] -1670850697
##
## $parameters$kinds

112
## [1] "Super-Duper"
##
##
## $statistics
## lambda treatmeans blockSize blocks r Efficiency
## values 3 9 3 36 12 0.75
##
## $sketch
## [,1] [,2] [,3]
## [1,] "8" "1" "3"
## [2,] "5" "3" "4"
## [3,] "3" "6" "4"
## [4,] "4" "9" "6"
## [5,] "7" "5" "8"
## [6,] "8" "4" "7"
## [7,] "1" "3" "5"
## [8,] "7" "4" "1"
## [9,] "2" "4" "1"
## [10,] "2" "7" "4"
## [11,] "3" "2" "7"
## [12,] "8" "1" "6"
## [13,] "7" "2" "8"
## [14,] "5" "7" "6"
## [15,] "1" "9" "7"
## [16,] "1" "3" "5"
## [17,] "8" "1" "9"
## [18,] "2" "9" "5"
## [19,] "6" "1" "7"
## [20,] "2" "1" "6"
## [21,] "6" "3" "9"
## [22,] "2" "9" "4"
## [23,] "4" "8" "5"
## [24,] "9" "6" "8"
## [25,] "8" "5" "9"
## [26,] "6" "2" "5"
## [27,] "9" "7" "3"
## [28,] "8" "4" "3"
## [29,] "2" "1" "5"
## [30,] "2" "8" "6"
## [31,] "3" "8" "2"
## [32,] "3" "6" "7"
## [33,] "4" "9" "1"
## [34,] "7" "9" "5"
## [35,] "4" "5" "6"
## [36,] "2" "9" "3"
##
## $book
## plots block treatments
## 1 101 1 8
## 2 102 1 1
## 3 103 1 3
## 4 201 2 5
## 5 202 2 3
## 6 203 2 4

113
## 7 301 3 3
## 8 302 3 6
## 9 303 3 4
## 10 401 4 4
## 11 402 4 9
## 12 403 4 6
## 13 501 5 7
## 14 502 5 5
## 15 503 5 8
## 16 601 6 8
## 17 602 6 4
## 18 603 6 7
## 19 701 7 1
## 20 702 7 3
## 21 703 7 5
## 22 801 8 7
## 23 802 8 4
## 24 803 8 1
## 25 901 9 2
## 26 902 9 4
## 27 903 9 1
## 28 1001 10 2
## 29 1002 10 7
## 30 1003 10 4
## 31 1101 11 3
## 32 1102 11 2
## 33 1103 11 7
## 34 1201 12 8
## 35 1202 12 1
## 36 1203 12 6
## 37 1301 13 7
## 38 1302 13 2
## 39 1303 13 8
## 40 1401 14 5
## 41 1402 14 7
## 42 1403 14 6
## 43 1501 15 1
## 44 1502 15 9
## 45 1503 15 7
## 46 1601 16 1
## 47 1602 16 3
## 48 1603 16 5
## 49 1701 17 8
## 50 1702 17 1
## 51 1703 17 9
## 52 1801 18 2
## 53 1802 18 9
## 54 1803 18 5
## 55 1901 19 6
## 56 1902 19 1
## 57 1903 19 7
## 58 2001 20 2
## 59 2002 20 1
## 60 2003 20 6

114
## 61 2101 21 6
## 62 2102 21 3
## 63 2103 21 9
## 64 2201 22 2
## 65 2202 22 9
## 66 2203 22 4
## 67 2301 23 4
## 68 2302 23 8
## 69 2303 23 5
## 70 2401 24 9
## 71 2402 24 6
## 72 2403 24 8
## 73 2501 25 8
## 74 2502 25 5
## 75 2503 25 9
## 76 2601 26 6
## 77 2602 26 2
## 78 2603 26 5
## 79 2701 27 9
## 80 2702 27 7
## 81 2703 27 3
## 82 2801 28 8
## 83 2802 28 4
## 84 2803 28 3
## 85 2901 29 2
## 86 2902 29 1
## 87 2903 29 5
## 88 3001 30 2
## 89 3002 30 8
## 90 3003 30 6
## 91 3101 31 3
## 92 3102 31 8
## 93 3103 31 2
## 94 3201 32 3
## 95 3202 32 6
## 96 3203 32 7
## 97 3301 33 4
## 98 3302 33 9
## 99 3303 33 1
## 100 3401 34 7
## 101 3402 34 9
## 102 3403 34 5
## 103 3501 35 4
## 104 3502 35 5
## 105 3503 35 6
## 106 3601 36 2
## 107 3602 36 9
## 108 3603 36 3

Example 15.3

115
example15.3 = data.frame(block = rep(c("1","2","3","4"),each=3),
treatment = c("T1","T3","T2",
"T4","T2","T1",
"T2","T3","T4",
"T3","T1","T4"),
height = c(14,12,13,
7.3,12.5,15.5,
12.8,15.5,11,
12.5,18.3,9.3))

model15.3 = aov(height ~ block+treatment,data=example15.3)

anova(model15.3)

## Analysis of Variance Table

##
## Response: height
## Df Sum Sq Mean Sq F value Pr(>F)
## block 3 4.556 1.5186 0.710 0.586366
## treatment 3 77.619 25.8731 12.097 0.009934 **
## Residuals 5 10.694 2.1388
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
HSD.test(model15.3,"treatment",console=TRUE)

##
## Study: model15.3 ~ "treatment"
##
## HSD Test for height
##
## Mean Square Error: 2.138833
##
## treatment, means
##
## height std r Min Max
## T1 15.93333 2.1825062 3 14.0 18.3
## T2 12.76667 0.2516611 3 12.5 13.0
## T3 13.33333 1.8929694 3 12.0 15.5
## T4 9.20000 1.8520259 3 7.3 11.0
##
## Alpha: 0.05 ; DF Error: 5
## Critical Value of Studentized Range: 5.218325
##
## Minimun Significant Difference: 4.406147
##
## Treatments with the same letter are not significantly different.
##
## height groups
## T1 15.93333 a
## T3 13.33333 ab
## T2 12.76667 ab
## T4 9.20000 b

116
Chapter 16: Split Plot Designs

Randomization

treatment1 = c("A","B","C","D")
treatment2 = c("I","II","III")
field_design = design.split(trt1 = treatment1, trt2 = treatment2,r = 3)
field_design$book

## plots splots block treatment1 treatment2

## 1 101 1 1 B I
## 2 101 2 1 B II
## 3 101 3 1 B III
## 4 102 1 1 C III
## 5 102 2 1 C II
## 6 102 3 1 C I
## 7 103 1 1 D III
## 8 103 2 1 D II
## 9 103 3 1 D I
## 10 104 1 1 A III
## 11 104 2 1 A I
## 12 104 3 1 A II
## 13 105 1 2 A I
## 14 105 2 2 A III
## 15 105 3 2 A II
## 16 106 1 2 B III
## 17 106 2 2 B II
## 18 106 3 2 B I
## 19 107 1 2 D II
## 20 107 2 2 D III
## 21 107 3 2 D I
## 22 108 1 2 C I
## 23 108 2 2 C II
## 24 108 3 2 C III
## 25 109 1 3 B II
## 26 109 2 3 B III
## 27 109 3 3 B I
## 28 110 1 3 D III
## 29 110 2 3 D I
## 30 110 3 3 D II
## 31 111 1 3 C II
## 32 111 2 3 C III
## 33 111 3 3 C I
## 34 112 1 3 A III
## 35 112 2 3 A I
## 36 112 3 3 A II

Example 16.2

example16.2 = data.frame(block=rep(c("1","2","3","4"),each=6),
factorA = c("A2","A2","A2","A1","A1","A1",
"A1","A1","A1","A2","A2","A2",

117
 "A1","A1","A1","A2","A2","A2",
"A2","A2","A2","A1","A1","A1"),
factorB = c("B2","B1","B3","B3","B2","B1",
"B1","B2","B3","B1","B3","B2",
"B3","B2","B1","B2","B3","B1",
"B3","B2","B1","B2","B3","B1"),
yield = c(25.5,24.9,25.8,26.1,18.0,21.7,
21.1,17.9,28.9,27.6,29.4,29.3,
28.6,19.5,23.2,29.7,30.3,29.5,
29.3,29.2,29.8,21.3,31.0,25.8))

In order to correctly split the stratums, we create the plotA variable which represents the plot units for
factor A.
example16.2$plotA = rep(c("1","2","3","4","5","6","7","8"),each=3)

model16.2 = aov(yield~block+factorA*factorB+Error(plotA),data=example16.2)
summary(model16.2)

##
## Error: plotA
## Df Sum Sq Mean Sq F value Pr(>F)
## block 3 55.06 18.35 5.158 0.10550
## factorA 1 136.33 136.33 38.318 0.00849 **
## Residuals 3 10.67 3.56
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Error: Within
## Df Sum Sq Mean Sq F value Pr(>F)
## factorB 2 98.37 49.18 102.23 2.90e-08 ***
## factorA:factorB 2 84.80 42.40 88.13 6.71e-08 ***
## Residuals 12 5.77 0.48
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
The pairwise test should be by stratum as follows and setting the degrees of freedom in the function.
HSD.test(example16.2$yield,example16.2$factorA,MSerror=1,DFerror=3,console=TRUE)

##
## Study: example16.2$yield ~ example16.2$factorA
##
## HSD Test for example16.2$yield
##
## Mean Square Error: 1
##
## example16.2$factorA, means
##
## example16.2.yield std r Min Max
## A1 23.59167 4.420090 12 17.9 31.0
## A2 28.35833 1.901415 12 24.9 30.3
##
## Alpha: 0.05 ; DF Error: 3
## Critical Value of Studentized Range: 4.500659
##

118
## Minimun Significant Difference: 1.299228
##
## Treatments with the same letter are not significantly different.
##
## example16.2$yield groups
## A2 28.35833 a
## A1 23.59167 b
HSD.test(example16.2$yield,example16.2$factorB,MSerror=2,DFerror=12,console=TRUE)

##
## Study: example16.2$yield ~ example16.2$factorB
##
## HSD Test for example16.2$yield
##
## Mean Square Error: 2
##
## example16.2$factorB, means
##
## example16.2.yield std r Min Max
## B1 25.450 3.347067 8 21.1 29.8
## B2 23.800 5.213992 8 17.9 29.7
## B3 28.675 1.848358 8 25.8 31.0
##
## Alpha: 0.05 ; DF Error: 12
## Critical Value of Studentized Range: 3.772929
##
## Minimun Significant Difference: 1.886464
##
## Treatments with the same letter are not significantly different.
##
## example16.2$yield groups
## B3 28.675 a
## B1 25.450 b
## B2 23.800 b
HSD.test(example16.2$yield,paste0(example16.2$factorB,example16.2$factorA),MSerror=2,DFerror=12,console=

##
## Study: example16.2$yield ~ paste0(example16.2$factorB, example16.2$factorA)
##
## HSD Test for example16.2$yield
##
## Mean Square Error: 2
##
## paste0(example16.2$factorB, example16.2$factorA), means
##
## example16.2.yield std r Min Max
## B1A1 22.950 2.095233 4 21.1 25.8
## B1A2 27.950 2.254625 4 24.9 29.8
## B2A1 19.175 1.594522 4 17.9 21.3
## B2A2 28.425 1.961929 4 25.5 29.7
## B3A1 28.650 2.007486 4 26.1 31.0
## B3A2 28.700 1.984943 4 25.8 30.3
##

119
## Alpha: 0.05 ; DF Error: 12
## Critical Value of Studentized Range: 4.750231
##
## Minimun Significant Difference: 3.358921
##
## Treatments with the same letter are not significantly different.
##
## example16.2$yield groups
## B3A2 28.700 a
## B3A1 28.650 a
## B2A2 28.425 a
## B1A2 27.950 a
## B1A1 22.950 b
## B2A1 19.175 c

120