Cap 11 Doane

A PowerPoint Presentation Package to Accompany
Applied Statistics in Business &

Economics, 6th edition
David P. Doane and Lori E. Seward
Prepared by Lloyd R. Jaisingh
Copyright ©2019 McGraw-Hill Education. All rights reserved. No reproduction or distribution without the
prior written consent of McGraw-Hill Education. 11-1
Chapter 11
Analysis of Variance
Chapter Contents
11.1 Overview of ANOVA

11.2 One-Factor ANOVA (Completely Randomized Model)
11.3 Multiple Comparisons
11.4 Tests for Homogeneity of Variances
11.5 Two-Factor ANOVA without Replication (Randomized Block
Model)
11.6 Two-Factor ANOVA with Replication (Full Factorial Model)
11.7 Higher Order ANOVA Models (Optional)
Chapter 11
Chapter Learning Objectives (LOs)
LO11-1: Use basic ANOVA terminology correctly.

LO11-2: Explain the assumptions of ANOVA and why they are important.
LO11-3: Recognize from data format when one-factor ANOVA is
appropriate.
LO11-4: Interpret sums of squares and calculations in an ANOVA table.
LO11-5: Use Excel or other software for ANOVA calculations.
LO11-6: Use a table or Excel to find critical values for the F distribution.
LO11-7: Understand and perform Tukey’s test for paired means.
Chapter 11
Chapter Learning Objectives (LOs), continued
LO11-8: Use Hartley’s test for equal variances in c treatment

groups.
LO11-9: Recognize from data format when two-factor ANOVA is
needed.
LO11-10: Interpret results in a two-factor ANOVA without replication.
LO11-11: Interpret main effects and interaction effects in two-factor
ANOVA.
LO11-12: Recognize the need for experimental design and GLM
(optional).
Chapter 11
11.1 Overview of ANOVA
LO11-1: Use basic ANOVA terminology correctly.
 Analysis of variance (ANOVA) is a comparison of

means.
 ANOVA allows one to compare more than two
means simultaneously.
 ANOVA allows one to trace sources of variation to
potential explanatory factors.
 Proper experimental design can make efficient use
of limited data to draw the strongest possible
inferences.
Chapter 11
LO11-1: Use basic ANOVA terminology correctly
(continued).
The Goal: Explaining Variation
• ANOVA seeks to identify sources of variation in a
numerical dependent variable Y (the response
variable).
• Variation in Y about its mean is explained by one or
more categorical independent variables (the factors) or
is unexplained (random error).
Chapter 11
(continued, 2).
The Goal: Explaining Variation

• Each possible value of a factor or combination of factors is a
treatment.
• We test to see if each factor has a significant effect on Y using
(for example) the hypotheses:
H0: μ1 = μ2 = μ3 = μ4 (e.g., mean defect rates are the same for
all four plants)
H1: Not all the means are equal.
• The test uses the F distribution.
• If we cannot reject H0, we conclude that observations within each
treatment have a common mean μ.
Chapter 11
(continued, 3).
Illustration: Manufacturing Defect Rates
• Figure 11.1 (on next slide) shows a dot plot of daily defect rates
for automotive computer chips manufactured at four plant
locations. Samples of 10 days’ production were taken at each
plant.
• Are the observed differences in the plants’ sample mean defect
rates merely due to random variation?
• Or are the observed differences between the plants’ defect
rates too great to be attributed to chance?
• This is the kind of question that one-factor ANOVA is designed
to answer.
Chapter 11
(continued, 4).
Illustration: Manufacturing Defect Rates (continued)
Chapter 11
(continued, 5).
A simple way to state the one-factor ANOVA hypothesis is
H0: μ1 = μ2 = μ3 = μ4 (mean defect rates are the same at all four plants)
H1: Not all the means are equal (at least one mean differs from the others)
If we cannot reject H0, then we conclude that the observations within each
treatment or group actually have a common mean μ (represented by a dashed
line in Figure 11.1).
This one-factor ANOVA model may be visualized as in Figure 11.2 as shown on

the next slide.
Chapter 11
(continued, 6).
Chapter 11
LO11-2: Explain the assumptions of ANOVA and why they
are important.
ANOVA Assumptions
• Analysis of Variance assumes that the

• observations on Y are independent,
• populations being sampled are normal,
• populations being sampled have equal variances.
• ANOVA is somewhat robust to departures from
normality and equal variance assumptions.
Chapter 11
LO11-2: Explain the assumptions of ANOVA and why they
are important (continued).
ANOVA Calculations
• Software (e.g., Excel, MegaStat, MINITAB, SPSS) can be

used to analyze data because computations are tedious.
• Large samples increase the power of the test,
but power also depends on the degree of variation in Y.
• Lowest power would be in a small sample with high
variation in Y.
Chapter 11
11.2 One-Factor ANOVA (Completely
Randomized Model)
LO11-3: Recognize from data format when one-factor ANOVA
is appropriate.
Data Format
 A one-factor ANOVA only compares the means of c groups
(treatments or factor levels).
 Consider the format for a one-factor ANOVA with treatments T1,
T2, …, Tc. Refer to the next slide.
Chapter 11
LO11-3: Recognize from data format when one-factor
ANOVA is appropriate (continued).
Data Format (continued)
 A one-factor ANOVA only compares the means of c groups

(treatments or factor levels).
 Consider the format for a one-factor ANOVA with treatments T1,
T2, …, Tc.
Chapter 11
ANOVA is appropriate (continued, 2).
Data Format (continued)
 Sample sizes within each treatment do not need to be equal (i.e.,
balanced).
 The total number of observations is equal to n = n1 + n2 + … + nc
Hypothesis to Be Tested
• ANOVA tests all means simultaneously and so does not inflate

the type I error.
Chapter 11
One-Factor ANOVA as a Linear Model

 An equivalent way to express the one-factor model is
to say that treatment j came from a population with a
common mean (μ) plus a treatment effect (Tj) plus
random error (eij):
 yij = m + Tj + eij
j = 1, 2, …, c and i = 1, 2, …, n
 Random error is assumed to be normally distributed

with zero mean and the same variance for all
treatments.
Chapter 11
One-Factor ANOVA as a Linear Model (continued)
 A fixed effects model only looks at what happens to the

response for particular levels of the factor.
H0: T1 = T2 = … = Tc = 0
H1: Not all Tj are zero
 If the H0 is true, then the ANOVA model collapses to yij
= μ + eij
Chapter 11
LO11-4: Interpret sums of squares and calculations in
an ANOVA table.
Group Means
 The mean of each group is calculated as:
• The overall sample mean (grand mean) can be calculated as:
Chapter 11
an ANOVA table (continued).
Partitioned Sum of Squares
To understand the logic of ANOVA, consider that for a

given observation yij the following relationship must hold
(on the right-hand side we just add and subtract :
Chapter 11
an ANOVA table (continued, 2).
Partitioned Sum of Squares (continued)

This says that any deviation of an observation from the grand
mean may be expressed in two parts: the deviation of the
column mean (j) from the grand mean (), or between
treatments, and the deviation of the observation (yij) from its
own column mean (), or within treatments. We can show that
this relationship also holds for sums of squared deviations,
yielding the partitioned sum of squares:
Chapter 11
This important relationship may be expressed simply

as
SST = SSB + SSE (partitioned sum of squares)
Chapter 11
 SSB and SSE are used to test the hypothesis of equal

treatment means by dividing each sum of squares by it
degrees of freedom to adjust for group size.
 These ratios are called Mean Squares (MSA and
MSE).
 The resulting test statistic is F = MSA/MSE.
 These calculations can be arranged in a worksheet like
Table 11.2 shown on the next slide.
Chapter 11
Chapter 11
LO11-5: Use EXCEL or other software for ANOVA
calculations.
 The ANOVA calculations are mathematically simple but involve

tedious sums.
 One can use Excel’s one-factor ANOVA menu using Data
Analysis to analyze data.
Chapter 11
calculations (continued).
Test Statistic
 The ANOVA test statistic is an F test statistic.
 The F distribution describes the ratio of two variances.
 The F statistic is the ratio of the variance due to treatments (MSB)
to the variance due to error (MSE).
 When F is near zero, then there is little difference among

treatments and we would not expect to reject the null hypothesis
of equal treatment means.
Chapter 11
calculations (continued, 2).
Decision Rule
• F cannot be negative and has no upper limit.
• For ANOVA, the F test is a right-tailed test.
• Use Appendix F or Excel (or other appropriate software) to obtain
the critical value of F for a given a.
Chapter 11
Example: Carton Packing

A cosmetics manufacturer’s regional distribution center has four
workstations that are responsible for packing cartons for shipment
to small retailers. Each workstation is staffed by two workers. The
task involves assembling each order, placing it in a shipping
carton, inserting packing material, taping the carton, and placing a
computer-generated shipping label on each carton. Generally,
each station can pack 200 cartons a day, and often more.
However, there is variability, due to differences in orders, labels,
and cartons. Table 11.3 (on next slide) shows the number of
cartons packed per day during a recent week. Is the variation
among stations within the range attributable to chance, or do
these samples indicate actual differences in the means?
Chapter 11
Example: Carton Packing (continued)
Chapter 11

As a preliminary step, we plot the data to check for any time
pattern and just to visualize the data. We see some potential
differences in means, but no obvious time pattern (otherwise we
would have to consider observation order as a second factor).
Figure 11.6
Chapter 11
Step 1: State the Hypothesis

The hypotheses to be tested are
H0: μ1 = μ2 = μ3 = μ4 (the means are the same)
H1: Not all the means are equal (at least one mean is different)
Step 2: State the Decision Rule

There are c = 4 groups and n = 20 observations, so degrees of
freedom for the F test are
Numerator: df1 = c – 1 = 4 – 1 = 3 (between treatments, factor)
Denominator: df2 = n – c = 20 – 4 = 16 (within treatments, error)
We will use = .05 for the test. The 5 percent right-tail critical value
from Appendix F is F3.16 = 3.24. Instead of Appendix F we could use
Excel’s function =F.INV.RT(0.05,3,16), which yields F.05 =
3.238872. This decision rule is illustrated in Figure 11.8.
Chapter 11
LO11-6: Use a table or Excel to find critical values for the
F distribution.
Chapter 11
F distribution (continued).
Chapter 11
F distribution (continued, 2).
Example: Carton Packing (continued).

Step 4: Make the Decision
Since the test statistic F = 4.12 exceed the critical value F.05 = 3.24,
we can reject the hypothesis of equal means. Since Excel gives the
p-value, you don’t actually need Excel’s critical value. The p-value (p
= .024124) is less than the level of significance ( = .05), which
confirms that we should reject the hypothesis of equal treatment
means. The Excel function for the p-value is
=F.DIST.RT(4.121769,3,16). For comparison, Figure 11.10 shows
MegaStat’s ANOVA table for the same data. The results are the
same, although MegaStat rounds things off, highlights significant p-
values, and gives standard deviations instead of variances for each
treatment.
Chapter 11
11.3 Multiple Comparison Tests
LO11-7: Understand and perform Tukey's test for paired
means.
Tukey’s Test
 After rejecting the null hypothesis of equal mean, we naturally
want to know: Which means differ significantly?
 In order to maintain the desired overall probability of type I error, a
simultaneous confidence interval for the difference of means must
be obtained.
 For c groups, there are c(c – 1)/2 distinct pairs of means to be
compared.
 These types of comparisons are called Multiple Comparison
Tests.
Chapter 11
means (continued).
Tukey’s Test (continued)
 Tukey’s studentized range test (or HSD for
“honestly significant difference” test) is a multiple
comparison test that has good power and is widely
used.
 Named for statistician John Wilder Tukey (1915 –
2000)
 This test is not available in Excel’s Tools > Data
Analysis but is available in most statistical
packages.
Chapter 11
means (continued, 2).

 Tukey’s is a two-tailed test for equality of paired means from c
groups compared simultaneously.
 The hypotheses are:
Decision Rule
Where Tc,n−c is a critical

value of the Tukey test
statistic Tcalc for the
desired level of
significance.
Chapter 11

 Note: If the desired degrees of freedom cannot be found, we
could interpolate or rely on a computer package to provide the
exact critical value. We take MSE directly from the ANOVA
calculations (see Table 11.2 in the text).
 The MSE is the pooled variance for all c samples combined
(rather than pooling just two sample variances as in Chapter 10).
Therefore, Tukey’s test statistic also could be written as
Chapter 11

 For example, here is the upper 5% of studentized range:
Chapter 11
11.4 Tests for Homogeneity of
Variances
LO11-8: Use Hartley’s test for equal variances in c treatment
groups.
ANOVA Assumptions
 ANOVA assumes that observations on the response variable are

from normally distributed populations that have the same
variance.
 The one-factor ANOVA test is only slightly affected by inequality
of variance when group sizes are equal.
 One can test this assumption of homogeneous variances by
using Hartley’s Fmax Test.
Chapter 11
LO11-8: Use Hartley's test for equal variances
in c treatment groups (continued).
Hartley’s Test
 The hypotheses are:
 The test statistic is the ratio of the largest sample variance to the
smallest sample variance.
𝑚𝑎𝑥 ( 𝑠1 , 𝑠 2 ,… , 𝑠𝑐 )
2 2 2 2
𝑠𝑚𝑎𝑥
𝐻 𝑐𝑎𝑙𝑐 = =
𝑚 𝑖𝑛 ( 𝑠1 , 𝑠 2 , … , 𝑠𝑐 )
2 2 2 2
𝑠 𝑚𝑖𝑛
Chapter 11
LO11-8: Use Hartley's test for equal variances in c
treatment groups (continued, 2).
Hartley’s Test (continued)
 The decision rule is:
Reject if
Critical values of may be found in Table 11.5 using

degrees of freedom given by
Numerator:
Denominator:
where n is the total number of observations. This test assumes
equal group sizes, so would be an integer. For group sizes that
are not drastically unequal, this procedure will still be
approximately correct, using the next lower integer if is not an
integer.
Chapter 11
Hartley’s Test (continued)
 Assuming equal
group sizes,
critical values of
Fmax are found
using degrees
of freedom
Chapter 11
Levene’s Test
 Levene’s test is a more robust alternative to Hartley’s

F test.
 Levene’s test does not assume a normal distribution.
 It is based on the distances of the observations from
their sample medians rather than their sample means.
 A computer program (e.g., MINITAB) is needed to
perform this test.
Chapter 11
Levene’s Test (continued)
for the
carton-packing data
Chapter 11
Please refer to your text for information
on Sections 11.5, 11.6 and 11.7

Cap 11 Doane

Uploaded by

Copyright:

Available Formats

Cap 11 Doane

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Cap 11 Doane

Uploaded by

Copyright:

Available Formats

A PowerPoint Presentation Package to Accompany

Applied Statistics in Business &

David P. Doane and Lori E. Seward

Prepared by Lloyd R. Jaisingh

11.1 Overview of ANOVA

LO11-1: Use basic ANOVA terminology correctly.

LO11-8: Use Hartley’s test for equal variances in c treatment

 Analysis of variance (ANOVA) is a comparison of

The Goal: Explaining Variation

A simple way to state the one-factor ANOVA hypothesis is

This one-factor ANOVA model may be visualized as in Figure 11.2 as shown on

• Analysis of Variance assumes that the

• Software (e.g., Excel, MegaStat, MINITAB, SPSS) can be

Data Format (continued)

 A one-factor ANOVA only compares the means of c groups

• ANOVA tests all means simultaneously and so does not inflate

One-Factor ANOVA as a Linear Model

 Random error is assumed to be normally distributed

One-Factor ANOVA as a Linear Model (continued)

 A fixed effects model only looks at what happens to the

 The mean of each group is calculated as:

• The overall sample mean (grand mean) can be calculated as:

Partitioned Sum of Squares

To understand the logic of ANOVA, consider that for a

Partitioned Sum of Squares (continued)

Partitioned Sum of Squares (continued)

This important relationship may be expressed simply

Partitioned Sum of Squares (continued)

 SSB and SSE are used to test the hypothesis of equal

Partitioned Sum of Squares (continued)

 The ANOVA calculations are mathematically simple but involve

 When F is near zero, then there is little difference among

Example: Carton Packing

Example: Carton Packing (continued)

Example: Carton Packing (continued)

Example: Carton Packing (continued)

Step 1: State the Hypothesis

Step 2: State the Decision Rule

Example: Carton Packing (continued)

Example: Carton Packing (continued)

Example: Carton Packing (continued).

Tukey’s Test (continued)

Where Tc,n−c is a critical

Tukey’s Test (continued)

Tukey’s Test (continued)

 ANOVA assumes that observations on the response variable are

Critical values of may be found in Table 11.5 using

 Levene’s test is a more robust alternative to Hartley’s

You might also like