STA 2 – LAB 6

Two-way factorial ANOVA

1. Objectives
- Check assumptions for two-way factorial ANOVA test.
- Conduct two-way factorial ANOVA test.
- Distinguish one-way and two-way ANOVA tests.
2. Procedure

We still base on codes used for one-way ANOVA to carry out two-way ANOVA test as follows:
 modelName <- aov(outcomevar~factor1*factor2,data=dataframe)#to see
variation by interaction
 summary(modelName)

Remember that this code just works in case all the sample sizes are equal. Besides, you must revise some
techniques to check the assumptions for ANOVA such as Q-Q plot, Levene’s Test for homogeneity of
variance, plot to see the validity of interaction between two factors, etc.

3. Exercises

Exercise 1. Does the frequency with which a supermarket product is offered at a discount affect the price
that customers expect to pay for the product? Does the percent reduction also affect this expectation? These
questions were examined by researchers in a study conducted on students enrolled in an introductory
management course at a large midwestern university. For 10 weeks 160 subjects received information about the
products. The treatment conditions corresponded to the number of promotions (1, 3, 5, or 7) during this 10-
week period and the percent at which the product was discounted (10%, 20%, 30%, and 40%). Ten students
were randomly assigned to each of the 4×4=16 treatments. For our case study we will examine the data for two
levels of promotions (1 and 5) and two levels of discount (10% and 30%). Data is stored in freqdisc2.csv file.

(a) Create a table summarising the sample size, mean, and standard deviation for each of the promotion-
by-discount combinations. Is it reasonable to pool the variances? Are normality assumptions satisfied?
(b) Run the analysis of variance. Report the F statistics with degrees of freedom and p-values for each of
the main effects and the interaction. What can you conclude? Write a short paragraph
summarizing the results of your analysis.

Import data from freqdisc2.csv file into R:

 freqdisc2 <-read.table("freqdisc2.csv", header=TRUE, sep = ",",

stringsAsFactors = FALSE)
 str(freqdisc2)

Because we want to see the combination promotions-by-discount, we must change variable Promotions
and Discount into factors:
 freqdisc2$Promotions <- factor(freqdisc2$Promotions, levels=c("1","5"),
labels=c("1 promotion","5 promotions"))
 freqdisc2$Discount <- factor(freqdisc2$Discount,levels = c("10","30"),
labels=c("10%","30%"))
A crosstabulation table between Promotions and Discount variables would give you the sample size
for each stratum.
1|P a ge
 STA 2 – LAB 6

 table(freqdisc2$Promotions,freqdisc2$Discount)

10% 30%
1promotion 10 10
5promotions 10 10

To describe mean and standard deviation of Price in terms of Promotions and Discount, use the code:
 by(freqdisc2$Price,list(freqdisc2$Promotions,freqdisc2$Discount), mean)

You will get the following output.

:1promotion
: 10%
[1] 4.92
----------------------------------------------
: 5promotions
: 10%
[1] 4.393
----------------------------------------------
: 1promotion
: 30%
[1] 4.225
----------------------------------------------
: 5promotions
: 30%
[1] 3.89

Below is the code to get the standard deviations for each sample:
 by(freqdisc2$Price,list(freqdisc2$Promotions,freqdisc2$Discount),sd)

: 1promotion
: 10%
[1] 0.1520234
----------------------------------------------
: 5promotions
: 10%
[1] 0.2685372
----------------------------------------------
: 1promotion
: 30%
[1] 0.3856092
----------------------------------------------
: 5promotions
: 30%
[1] 0.1628906

Next, we’re going to check the assumption of equal standard deviations. The ratio of largest SD over
smallest SD is around 2.54 (which is between 2 and 3 and in this case it is not so clear to pool variances), then
it’s good to check again using Levene’s test:
 leveneTest(freqdisc2$Price,interaction(freqdisc2$Promotions,
freqdisc2$Discount),center=median)

2|P a ge
 STA 2 – LAB 6

The test gives you:

Levene's Test for Homogeneity of Var
iance (center = median)
Df F value Pr(>F)
group 3 2.7878 0.05451 .
36
---
Signif. codes: 0 ‘***’ 0.001 ‘**’ 0
.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

What is your conclusion about the assumption

of equal standard deviations?
We check the assumption of normality using Q-
Q plot:
 library(car)
 qqPlot(lm(Price ~ Promotions +
Discount + Promotions*Discount, data=freqdisc2), simulate=T, main="Q-Q
Plot", labels=F)
What can you say about normality of residuals based on the above Q-Q plot?

The sample sizes of all groups are not so large (just 10 observations) then it’s not appropriate to use boxplots
to compare 4 groups, instead, we’d like to use meanplots. The codes and outputs are provided below:
 install.packages("gplots")
 library(gplots)
 plotmeans(Price ~ interaction(Promotions,Discount), data = freqdisc2, xla
b = "Promotions and Discount", ylab = "Expected prices", main="Mean Plot
with 95% CI")

3|P a ge
 STA 2 – LAB 6

Two-way ANOVA
Now we’re going to run two-way ANOVA test with Price as outcome variable and Promotions and
Discounts as two factors. We’re also interested in the main effects of Promotions and Discount and their
interaction, so we use the format Price ~ Promotions*Discount:
 freqdisc2.result<-aov(Price ~ Promotions*Discount, data = freqdisc2)
 summary(freqdisc2.result)

Here is the R output for two-way ANOVA test:

Df Sum Sq Mean Sq F value Pr(>F)
Promotions 1 1.858 1.858 27.474 7.17e-06 ***
Discount 1 3.588 3.588 53.067 1.39e-08 ***
Promotions:Discount 1 0.092 0.092 1.363 0.251
Residuals 36 2.434 0.068
---
Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01
‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

Questions: How do promotions and discount and their

interaction affect the expected price? Provide comments.
Interaction Plot:
We want to see the interaction between two factors
graphically, so we use the interaction.plot
function as follows.
 interaction.plot(freqdisc2$Promoti
ons, freqdisc2$Discount,
freqdisc2$Price, type="b", col=c("red",
"blue"), pch=c(16, 18), main =
"Interaction between Promotions and Discount")

Note that because the interaction effect is not significant, we do not interpret the interaction plot. In practice,
you do not need to produce an interaction plot if the interaction effect is not significant.

Exercise 2. A study of cardiovascular risk factors compared runners who averaged at least 15 miles per week
with a control group described as “generally sedentary.” Both men and women were included in the study. The
data set was constructed based on information provided in P. D. Wood et al., “Plasma lipoprotein distributions
in male and female runners,” in P. Milvey (ed.), The Marathon: Physiological, Medical, Epidemiological, and
Psychological Studies, New York Academy of Sciences, 1977. The study design is a 2×2 ANOVA with the
factors group and gender. There were 200 subjects in each of the four combinations. The variables are ID, a
numeric subject identifier; Group, with values “Control” and “Runners”; Gender, with values Female and
Male; and HeartRate, heart rate after the subject ran for six minutes on a treadmill. Analyze the data using a
two-way ANOVA. Summarize your findings in a short report. The data file is runners.csv.
1. Import data from runners.csv into R, then check the first 6 subjects (using head()) as well as
structure of this dataframe.

4|P a ge
 STA 2 – LAB 6

Id Group Gender HeartRate

1 1 Control Female 159
2 2 Control Female 183
3 3 Control Female 140
4 4 Control Female 140
5 5 Control Female 125
6 6 Control Female 155

'data.frame': 800 obs. of 4 variables:

$ Id : int 1 2 3 4 5 6 7 8 9 10 ...
$ Group : Factor w/ 2 levels "Control","Runners": 1 1 1 1 1 1 1 1 1 1 ...
$ Gender : Factor w/ 2 levels "Female","Male": 1 1 1 1 1 1 1 1 1 1 ...
$ HeartRate: int 159 183 140 140 125 155 148 132 158 136 ...
2. Crosstabulation table between 2 factors:
Female Male
Control 200 200
Runners 200 200
Graphical description:

3. Means for groups:

: Control
: Female
[1] 148
----------------------------------------------
: Runners
: Female
[1] 115.985
----------------------------------------------
: Control
: Male
[1] 130
----------------------------------------------

5|P a ge
 STA 2 – LAB 6

: Runners
: Male
[1] 103.975
4. Standard deviation for groups:
: Control
: Female
[1] 16.27095
----------------------------------------------
: Runners
: Female
[1] 15.97154
----------------------------------------------
: Control
: Male
[1] 17.10035
----------------------------------------------
: Runners
: Male
[1] 12.49942
5. Check the homogeneity of variances:
Check the assumption of equal standard deviations using the rule we learnt in the lecture. Be careful when you
use the Levene’s test for large sample sizes:

Levene's Test for Homogeneity of Variance (center = median)

Df F value Pr(>F)
group 3 5.7339 0.0006971 ***
796
---
Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

6. Check the normality of residuals:

6|P a ge
 STA 2 – LAB 6

7. Two-way ANOVA:
Df Sum Sq Mean Sq F value Pr(>F)
Group 1 168432 168432 695.647 < 2e-16 ***
Gender 1 45030 45030 185.980 < 2e-16 ***
Group:Gender 1 1794 1794 7.409 0.00663 **
Residuals 796 192730 242
---
Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

Note: If the interaction effect is significant, you should ignore the main effects.

Interaction Plot

7|P a ge