Chapter 6

Analysis of Variance
( 3 hours)
 Chapter outline

 One-Way Analysis of Variance

 Multiple Comparisons (optional)
 Concepts

The response variable is the variable of interest to

be measured in the experiment. We also refer to the
response as the dependent variable. Typically, the
response/dependent variable is quantitative in nature.
Factors are those variables whose effect on the
response is of interest to the experimenter.
Quantitative factors are measured on a numerical
scale, whereas qualitative factors are those that are
not (naturally) measured on a numerical scale.
Factors are also referred to as independent
variables.
 Concepts

Factor levels are the values of the factor used in

the experiment.

The treatments of an experiment are the

factor-level combinations used.
An experimental unit is the object on which the
response and factors are observed or measured.
 One-Way Analysis of
Variance
 The analysis of variance is a procedure that
tests to determine whether differences exist
between two or more population means.
 Example
 In the last decade, stockbrokers have drastically
changed the way they do business. Internet trading
has become quite common, and online trades can cost
as little as $7. It is now easier and cheaper to invest in
the stock market than ever before. What are the
effects of these changes? To help answer this question,
a financial analyst randomly sampled 366 American
households and asked each to report the age category
of the head of the household and the proportion of its
financial assets that are invested in the stock market.
The age categories are Young (less than 35),
Early middle age (35 to 49), Late middle age (50
to 65), Senior (older than 65).
 Example
 The analyst was particularly interested in
determining whether the ownership of
stocks varied by age. Some of the data are
listed next. Do these data allow the analyst
to determine that there are differences in
stock ownership between the four age
groups?
 Concepts

 The variable X is called the response variable,

and its values are called responses.
 The unit that we measure is called an
experimental unit. In this example, the
response variable is the percentage of assets
invested in stocks, and the experimental units are
the heads of households sampled.
 The criterion by which we classify the populations
is called a factor. Each population is called a
factor level. In this example, factor is is the age
category of the head of the household and there
are four levels.
 Test statistics
 If the null hypothesis is true, the population
means would all be equal. We would then expect
that the sample means would be close to one
another. If the alternative hypothesis is true,
however, there would be large differences
between some of the sample means.
 The statistic that measures the proximity of the
sample means to each other is called the
between-treatments variation; it is denoted
SST, which stands for sum of squares for
treatments.
 Test statistics
 If large differences exist between the sample means,
at least some sample means differ considerably
from the grand mean, producing a large value of
SST.
 How large does the statistic have to be for us to
justify rejecting the null hypothesis.
 Within-treatments variation, which is denoted by
SSE (sum of squares for error): a measure of the
amount of variation in the response variable that is
not caused by the treatments.
Test statistics
 ANOVA F-Test to Compare k
Treatment Means:
H0: µ1 = µ2 = … = µk
Ha: At least two treatment means differ
MST
Test Statistic: F 
MSE
Rejection region: F > F, where F is based on (k –
1) numerator degrees of freedom (associated with
MST) and (n – k) denominator degrees of freedom
(associated with MSE).
 ANOVA F-Test Critical Value

If means are equal, F

= MST / MSE  1.
Only reject large F! Reject H 0

Do Not 
Reject H 0

0 F
F(α; k – 1, n – k)

Always One-Tail!
© 1984-1994 T/Maker Co.
 Example
 You select independent random samples of five female
and five male high school seniors and record their SAT
scores Can we conclude that the population of female
high school students scores higher, on average, than the
population of male students?
Example
 Are the assumptions required for the
test approximately satisfied?

 The samples of golf balls for each brand are

selected randomly and independently.
 The probability distributions of the distances for
each brand are normal.
 The variances of the distance probability
distributions for each brand are equal.
 ANOVA Summary Table

Degrees Mean
Source of Sum of
of Square F
Variation Squares
Freedom (Variance)
Treatment k–1 SST MST = MST
SST/(k – 1) MSE
Error n–k SSE MSE =
SSE/(n – k)
Total n–1 SS(Total) =
SST+SSE
 ANOVA F-Test Hypotheses
 H0: 1 = 2 = 3 = ... = k
f(x)
— All population means
are equal
— No treatment effect
x
1 = 2 = 3
 Ha: Not All i Are Equal
— At least 2 pop. means
f(x)
are different
— Treatment effect

—    ...   is
1 2 k x
Wrong 1 =  2  3
 ANOVA F-Test Example
As production manager,
you want to see if three Mach1 Mach2 Mach3
filling machines have 25.40 23.40 20.00
different mean filling times. 26.31 21.80 22.20
You assign 15 similarly 24.10 23.50 19.75
trained and experienced 23.74 22.75 20.60
workers, 5 per machine, to 25.10 21.60 20.40
the machines. At the .05
level of significance, is
there a difference in mean
filling times?
 ANOVA F-Test
Thinking Challenge

You’re a trainer for Microsoft

Corp. Is there a difference in
mean learning times of 12
people using 4 different
training methods ( =.05)?
M1 M2 M3 M4
© 1984-1994 T/Maker Co.

10 11 13 18
9 16 8 23
5 9 9 25
Use the following table.
 Summary Table
(Partially Completed)

Degrees Mean
Source of Sum of
of Square F
Variation Squares (Variance)
Freedom
Treatment 348
(Methods)
Error 80

Total
The Randomized Block Design
 Randomized Block Design
The randomized block design consists of a two-
step procedure:
1. Matched sets of experimental units, called
blocks, are formed, each block consisting of k
experimental units (where k is the number of
treatments). The b blocks should consist of
experimental units that are as similar as possible.
2. One experimental unit from each block is
randomly assigned to each treatment, resulting in
a total of
n = bk responses.
 Randomized Block Design
 Reduces sampling variability (SSE)
 Matched sets of experimental units (blocks)
 One experimental unit from each block is
randomly assigned to each treatment
Randomized Block Design
Total Variation Partitioning
 ANOVA F-Test to Compare k Treatment
Means: Randomized Block Design

H0: µ1 = µ2 = … = µk
Ha: At least two treatment means differ
MST
Test Statistic: F 
MSE
Rejection region: F > F, where F is based on (k –
1) numerator degrees of freedom and (n – b – k
+1) denominator degrees of freedom.
 Conditions Required for a Valid ANOVA F-test:
Randomized Block Design

1. The b blocks are randomly selected, and all k

treatments are applied (in random order) to
each block.
2. The distributions of observations
corresponding to all bk block-treatment
combinations are approximately normal.
3. All bk block-treatment distributions have equal
variances.
 Randomized Block Design
F-Test Test Statistic
1. Test Statistic
•
F = MST / MSE
— MST is Mean Square for Treatment
— MSE is Mean Square for Error
2. Degrees of Freedom
• 1 = k – 1 (numerator) 2 = n – k – b + 1
(denominator)
— k = Number of groups
— n = Total sample size
— b = Number of blocks
 Randomized Block Design Example
A production manager wants to see if three
assembly methods have different mean assembly
times (in minutes). Five employees were selected
at random and assigned to use each assembly
method. At the .05 level of significance, is there a
difference in mean assembly times?
Employee Method 1 Method 2 Method
3
1 5.4 3.6 4.0
2 4.1 3.8 2.9
3 6.1 5.6 4.3
4 3.6 2.3 2.6
5 5.3 4.7 3.4
Steps for Conducting an ANOVA for
a Randomized Block Design
1. Be sure the design consists of blocks (preferably,
blocks of homogeneous experimental units) and
that each treatment is randomly assigned to one
experimental unit in each block.
2. If possible, check the assumptions of normality
and equal variances for all block-treatment
combinations. [Note: This may be difficult to do
because the design will likely have only one
observation for each block-treatment combination.]
Steps for Conducting an ANOVA for
a Randomized Block Design
3. Create an ANOVA summary table that specifies
the variability attributable to Treatments,
Blocks, and Error, which leads to the
calculation of the F-statistic to test the null
hypothesis that the treatment means are equal
in the population. Use a statistical software
package or the calculation formulas in
Appendix C to obtain the necessary numerical
ingredients.
Steps for Conducting an ANOVA
for a Randomized Block Design
4. If the F-test leads to the conclusion that the
means differ, use the Bonferroni, Tukey, or
similar procedure to conduct multiple
comparisons of as many of the pairs of means
as you wish. Use the results to summarize the
statistically significant differences among the
treatment means. Remember that, in general,
the randomized block design cannot be used to
form confidence intervals for individual treatment
means.
Steps for Conducting an ANOVA
for a Randomized Block Design
5. If the F-test leads to the nonrejection of the null
hypothesis that the treatment means are equal,
consider the following possibilities:
a. The treatment means are equal–that is, the null
hypothesis is true.
Steps for Conducting an ANOVA for
a Randomized Block Design
b. The treatment means really differ, but other
important factors affecting the response are not
accounted for by the randomized block design.
These factors inflate the sampling variability, as
measured by MSE, resulting in smaller values of
the F-statistic. Either increase the sample size for
each treatment or conduct an experiment that
accounts for the other factors affecting the
response. Do not automatically reach the former
conclusion because the possibility of a Type II
error must be considered if you accept H0.
Steps for Conducting an ANOVA
for a Randomized Block Design

6. If desired, conduct the F-test of the null

hypothesis that the block means are equal.
Rejection of this hypothesis lends statistical
support to using the randomized block design.
Steps for Conducting an ANOVA
for a Randomized Block Design

Note: It is often difficult to check whether the

assumptions for a randomized block design are
satisfied. There is usually only one observation for
each block-treatment combination. When you feel
these assumptions are likely to be violated, a
nonparametric procedure is advisable.
Factorial Experiments:
Two Factors
 Factorial Design

A complete factorial experiment is one in which

every factor-level combination is employed–that is,
the number of treatments in the experiment equals
the total number of factor-level combinations.
Also referred to as a two-way classification.
 Factorial Design
To determine the nature of the treatment effect, if
any, on the response in a factorial experiment, we
need to break the treatment variability into three
components: Interaction between Factors A and B,
Main Effect of Factor A, and Main Effect of Factor
B. The Factor Interaction component is used to
test whether the factors combine to affect the
response, while the Factor Main Effect
components are used to determine whether the
factors separately affect the response.
 Factorial Design

 Experimental units (subjects) are assigned

randomly to treatments
 Subjects are assumed homogeneous
 Two or more factors or independent variables
 Each has two or more treatments (levels)
 Analyzed by two-way ANOVA
 Procedure for Analysis of Two-
Factor Factorial Experiment

1. Partition the Total Sum of Squares into the

Treatments and Error components. Use either a
statistical software package or the calculation
formulas in Appendix C to accomplish the
partitioning.
 Procedure for Analysis of Two-
Factor Factorial Experiment

2. Use the F-ratio of Mean Square for Treatments to

Mean Square for Error to test the null hypothesis
that the treatment means are equal.
a. If the test results in nonrejection of the null
hypothesis, consider refining the experiment by
increasing the number of replications or introducing
other factors. Also consider the possibility that the
response is unrelated to the two factors.
b. If the test results in rejection of the null hypothesis,
then proceed to step 3.
 Procedure for Analysis of Two-
Factor Factorial Experiment

3. Partition the Treatments Sum of Squares into the

Main Effect and Interaction Sum of Squares. Use
either a statistical software package or the
calculation formulas in Appendix C to accomplish
the partitioning.
 Procedure for Analysis of Two-
Factor Factorial Experiment

4. Test the null hypothesis that factors A and B do

not interact to affect the response by computing
the F-ratio of the Mean Square for Interaction to
the Mean Square for Error.
a. If the test results in nonrejection of the null
hypothesis, proceed to step 5.
b. If the test results in rejection of the null hypothesis,
conclude that the two factors interact to affect the
mean response. Then proceed to step 6a.
 Procedure for Analysis of Two-
Factor Factorial Experiment

5. Conduct tests of two null hypotheses that the

mean response is the same at each level of factor
A and factor B. Compute two F-ratios by
comparing the Mean Square for each Factor Main
Effect to the Mean Square for Error.
a. If one or both tests result in rejection of the null
hypothesis, conclude that the factor affects the
mean response. Proceed to step 6b.
 Procedure for Analysis of Two-
Factor Factorial Experiment

b. If both tests result in nonrejection, an apparent

contradiction has occurred. Although the
treatment means apparently differ (step 2 test),
the interaction (step 4) and main effect (step 5)
tests have not supported that result. Further
experimentation is advised.
 Procedure for Analysis of Two-
Factor Factorial Experiment

6. Compare the means:

a. If the test for interaction (step 4) is significant, use
a multiple comparisons procedure to compare any
or all pairs of the treatment means.
b. If the test for one or both main effects (step 5) is
significant, use a multiple comparisons procedure
to compare the pairs of means corresponding to
the levels of the significant factor(s).
ANOVA Tests Conducted for Factorial Experiments:
Completely Randomized Design, r Replicates per
Treatment

Test for Treatment Means

H0: No difference among the ab treatment means
Ha: At least two treatment means differ
MST
Test Statistic: F 
MSE
Rejection region: F > F, based on (ab – 1)
numerator and (n – ab) denominator degrees of
freedom [Note: n = abr.]
 ANOVA Tests Conducted for Factorial Experiments:
Completely Randomized Design, r Replicates per
Treatment

Test for Factor Interaction

H0: Factors A and B do not interact to affect the
response mean
Ha: Factors A and B do interact to affect the response
mean
MS  AB 
Test Statistic: F 
MSE
Rejection region: F > F, based on (a – 1)(b – 1)
numerator and (n – ab) denominator degrees of freedom
ANOVA Tests Conducted for Factorial Experiments:
Completely Randomized Design, r Replicates per
Treatment

Test for Main Effect of Factor A

H0: No difference among the a mean levels of factor
A
Ha: At least two factor A mean levels differ
MS  A 
Test Statistic: F 
MSE
Rejection region: F > F, based on (a – 1)
numerator and (n – ab) denominator degrees of
freedom
ANOVA Tests Conducted for Factorial Experiments:
Completely Randomized Design, r Replicates per
Treatment

Test for Main Effect of Factor B

H0: No difference among the b mean levels of factor
B
Ha: At least two factor B mean levels differ
Test Statistic: F 

MS B
MSE
Rejection region: F > F, based on (b – 1)
numerator and (n – ab) denominator degrees of
freedom
 Conditions Required for Valid
F-tests in Factorial Experiments

1. The response distribution for each factor-level

combination (treatment) is normal.
2. The response variance is constant for all
treatments.
3. Random and independent samples of
experimental units are associated with each
treatment.
 ANOVA Data Table

Factor Factor B
A 1 2 ... b Observation k
1 x111 x121 ... x1b1
x112 x122 ... x1b2 xijk
2 x211 x221 ... x2b1
x212 x222 ... x2b2 Level i Level j
: : : : : Factor Factor
a xa11 xa21 ... xab1 A B

xa12 xa22 ... xab2 Treatment

 Factorial Design Example

Factor 2 (Training Method)

Factor Level 1 Level 2 Level 3
Levels

Level 1 15 hr.  10 hr.  22 hr. 

(High)
Factor 1 11 hr.  12 hr.  17 hr. 
(Motivation)
Level 2 27 hr.  15 hr.  31 hr. 
(Low)
29 hr.  17 hr.  49 hr. 
Treatment
 Advantages
of Factorial Designs

 Saves time and effort

 e.g., Could use separate completely
randomized designs for each variable
 Controls confounding effects by putting other
variables into model
 Can explore interaction between variables
 Two-Way ANOVA

 Tests the equality of two or more population

means when several independent variables
are used
 Same results as separate one-way ANOVA on
each variable
 No interaction can be tested
 Used to analyze factorial designs
 Interaction

• Occurs when effects of one factor vary

according to levels of other factor
• When significant, interpretation of main effects
(A and B) is complicated
• Can be detected
– In data table, pattern of cell means in one row
differs from another row
– In graph of cell means, lines cross
 Graphs of Interaction
Effects of motivation (high or low) and
training method (A, B, C) on mean learning
time
Interaction No Interaction
Average Average
Response High Response High

Low Low

A B C A B C
 Two-Way ANOVA
Total Variation Partitioning
 Two-Way ANOVA
Summary Table

Source of Degrees of Sum of Mean F

Variation Freedom Squares Square
A a–1 SS(A) MS(A) MS(A)
(Row) MSE
B b–1 SS(B) MS(B) MS(B)
(Column) MSE
AB (a – 1)(b – 1) SS(AB) MS(AB) MS(AB)
(Interaction) MSE
Error n – ab SSE MSE
Total n–1 SS(Total) Same as other
designs
 Factorial Design
Example
Human Resources wants to determine if training
time is different based on motivation level and
training method. Conduct the appropriate ANOVA
tests. Use α = .05 for each test.
Training Method
Factor Self–
Levels paced Classroom Computer
15 hr. 10 hr. 22 hr.
High
11 hr. 12 hr. 17 hr.
Motivation
27 hr. 15 hr. 31 hr.
Low
29 hr. 17 hr. 49 hr.
 End of chapter 6