Slides by
JOHN
LOUCKS
& Updated by
SPIROS
VELIANITIS
2008 Thomson South-Western. All Rights Reserved Slide 1
Chapter 13
Experimental Design and Analysis of Variance
Introduction to Experimental Design
and Analysis of Variance
Analysis of Variance
and the Completely Randomized Design
Multiple Comparison Procedures
2008 Thomson South-Western. All Rights Reserved Slide 2
An Introduction to Experimental Design
and Analysis of Variance
Statistical studies can be classified as being either
experimental or observational.
In an experimental study, one or more factors are
controlled so that data can be obtained about how the
factors influence the variables of interest.
In an observational study, no attempt is made to
control the factors.
Cause-and-effect relationships are easier to establish
in experimental studies than in observational studies.
Analysis of variance (ANOVA) can be used to analyze
the data obtained from experimental or observational
studies.
2008 Thomson South-Western. All Rights Reserved Slide 3
An Introduction to Experimental Design
and Analysis of Variance
In this chapter three types of experimental designs
are introduced.
a completely randomized design
a randomized block design
a factorial experiment
2008 Thomson South-Western. All Rights Reserved Slide 4
An Introduction to Experimental Design
and Analysis of Variance
A factor is a variable that the experimenter has
selected for investigation (the independent variable).
A treatment is a level of a factor.
Experimental units are the objects of interest in the
experiment.
A completely randomized design is an experimental
design in which the treatments are randomly
assigned to the experimental units.
2008 Thomson South-Western. All Rights Reserved Slide 5
Analysis of Variance: A Conceptual Overview
Analysis of Variance (ANOVA) can be used to test
for the equality of three or more population means.
Data obtained from observational or experimental
studies can be used for the analysis.
We want to use the sample results to test the
following hypotheses:
H0: 1 = 2 = 3 = . . . = k
Ha: Not all population means are equal
2008 Thomson South-Western. All Rights Reserved Slide 6
Analysis of Variance: A Conceptual Overview
H0: 1 = 2 = 3 = . . . = k
Ha: Not all population means are equal
If H0 is rejected, we cannot conclude that all
population means are different.
Rejecting H0 means that at least two population
means have different values.
2008 Thomson South-Western. All Rights Reserved Slide 7
Analysis of Variance: A Conceptual Overview
Assumptions for Analysis of Variance
For each population, the response (dependent)
variable is normally distributed.
The variance of the response variable, denoted 2,
is the same for all of the populations.
The observations must be independent.
2008 Thomson South-Western. All Rights Reserved Slide 8
Analysis of Variance: A Conceptual Overview
Sampling Distribution of x Given H0 is True
Sample means are close together
because there is only
one sampling distribution
when H0 is true.
2
x2
n
x2 x1 x3
2008 Thomson South-Western. All Rights Reserved Slide 9
Analysis of Variance: A Conceptual Overview
Sampling Distribution of x Given H0 is False
Sample means come from
different sampling distributions
and are not as close together
when H0 is false.
x3 3 x1 1 2 x2
2008 Thomson South-Western. All Rights Reserved Slide 10
Analysis of Variance
Between-Treatments Estimate of Population Variance
Within-Treatments Estimate of Population Variance
Comparing the Variance Estimates: The F Test
ANOVA Table
2008 Thomson South-Western. All Rights Reserved Slide 11
Between-Treatments Estimate
of Population Variance 2
The estimate of 2 based on the variation of the
sample means is called the mean square due to
treatments and is denoted by MSTR.
k
j j
n (
j 1
x x ) 2
MSTR
k1
Denominator is the Numerator is called
degrees of freedom the sum of squares due
associated with SSTR to treatments (SSTR)
2008 Thomson South-Western. All Rights Reserved Slide 12
Within-Treatments Estimate
of Population Variance 2
The estimate of 2 based on the variation of the
sample observations within each sample is called the
mean square error and is denoted by MSE.
k
2
(n j 1) s 2j
j1
MSE
nT k
Numerator is called
Denominator is the the sum of squares
degrees of freedom due to error (SSE)
associated with SSE
2008 Thomson South-Western. All Rights Reserved Slide 13
Comparing the Variance Estimates: The F Test
If the null hypothesis is true and the ANOVA
assumptions are valid, the sampling distribution of
MSTR/MSE is an F distribution with MSTR d.f.
equal to k - 1 and MSE d.f. equal to nT - k.
If the means of the k populations are not equal, the
value of MSTR/MSE will be inflated because MSTR
overestimates 2.
Hence, we will reject H0 if the resulting value of
MSTR/MSE appears to be too large to have been
selected at random from the appropriate F
distribution.
2008 Thomson South-Western. All Rights Reserved Slide 14
Comparing the Variance Estimates: The F Test
Sampling Distribution of MSTR/MSE
Sampling Distribution
of MSTR/MSE
Reject H0
Do Not Reject H0
MSTR/MSE
F
Critical Value
2008 Thomson South-Western. All Rights Reserved Slide 15
Test for the Equality of k Population Means
Hypotheses
H0: 1 = 2 = 3 = . . . = k
Ha: Not all population means are equal
Test Statistic
F = MSTR/MSE
2008 Thomson South-Western. All Rights Reserved Slide 16
Test for the Equality of k Population Means
Rejection Rule
p-value Approach: Reject H0 if p-value <
Critical Value Approach: Reject H0 if F > F
where the value of F is based on an
F distribution with k - 1 numerator d.f.
and nT - k denominator d.f.
2008 Thomson South-Western. All Rights Reserved Slide 17
ANOVA Table
The ANOVA table decomposes the variance into two
components (sources): a between-group component
(treatments) and a within-group component (error). The F-
ratio, which in this case equals 15.6234, is a ratio of the
between-group estimate to the within-group estimate. Since
the P-value of the F-test is less than 0.05, there is a statistically
significant difference between the means from one level to
another at the 95.0% confidence level.
2008 Thomson South-Western. All Rights Reserved Slide 18
Testing for the Equality of k Population Means:
A Completely Randomized Experimental Design
Example: AutoShine, Inc.
AutoShine, Inc. is considering marketing a long-
lasting car wax. Three different waxes (Type 1, Type 2,
and Type 3) have been developed.
In order to test the durability
of these waxes, 5 new cars were
waxed with Type 1, 5 with Type
2, and 5 with Type 3. Each car was then
repeatedly run through an automatic carwash until the
wax coating showed signs of deterioration.
2008 Thomson South-Western. All Rights Reserved Slide 19
Testing for the Equality of k Population Means:
A Completely Randomized Experimental Design
Example: AutoShine, Inc.
The number of times each car went through the
carwash before its wax deteriorated is shown on the
next slide. AutoShine, Inc. must decide which wax
to market. Are the three waxes
equally effective?
Factor . . . Car wax
Treatments . . . Type I, Type 2, Type 3
Experimental units . . . Cars
Response variable . . . Number of washes
2008 Thomson South-Western. All Rights Reserved Slide 20
Testing for the Equality of k Population Means:
A Completely Randomized Experimental Design
Wax Wax Wax
Observation Type 1 Type 2 Type 3
1 27 33 29
2 30 28 28
3 29 31 30
4 28 30 32
5 31 30 31
Sample Mean 29.0 30.4 30.0
Sample Variance 2.5 3.3 2.5
2008 Thomson South-Western. All Rights Reserved Slide 21
Testing for the Equality of k Population Means:
A Completely Randomized Experimental Design
Hypotheses
H0: 1 = 2 = 3
Ha: Not all the means are equal
where:
1 = mean number of washes using Type 1 wax
2 = mean number of washes using Type 2 wax
3 = mean number of washes using Type 3 wax
2008 Thomson South-Western. All Rights Reserved Slide 22
Testing for the Equality of k Population Means:
A Completely Randomized Experimental Design
Test Statistic
F = MSTR/MSE
Conclusion
The p-value of 0.42 is greater than 0.10, where F = 2.81.
Therefore, we cannot reject H0.
There is insufficient evidence to conclude that
the mean number of washes for the three wax
types are not all the same (waxes are the same).
2008 Thomson South-Western. All Rights Reserved Slide 23
Testing for the Equality of k Population Means:
A Completely Randomized Experimental Design
ANOVA Table
Source of Sum of Degrees of Mean
Variation Squares Freedom Squares F p-Value
Treatments 5.2 2 2.60 .939 .42
Error 33.2 12 2.77
Total 38.4 14
2008 Thomson South-Western. All Rights Reserved Slide 24
Testing for the Equality of k Population Means:
An Observational Study
Example: Reed Manufacturing
Janet Reed would like to know if
there is any significant difference in
the mean number of hours worked per
week for the department managers
at her three manufacturing plants
(in Buffalo, Pittsburgh, and Detroit).
An F test will be conducted using
= .05.
2008 Thomson South-Western. All Rights Reserved Slide 25
Testing for the Equality of k Population Means:
An Observational Study
Example: Reed Manufacturing
A simple random sample of five
managers from each of the three plants
was taken and the number of hours
worked by each manager in the
previous week is shown on the next
slide.
Factor . . . Manufacturing plant
Treatments . . . Buffalo, Pittsburgh, Detroit
Experimental units . . . Managers
Response variable . . . Number of hours worked
2008 Thomson South-Western. All Rights Reserved Slide 26
Testing for the Equality of k Population Means:
An Observational Study
Plant 1 Plant 2 Plant 3
Observation Buffalo Pittsburgh Detroit
1 48 73 51
2 54 63 63
3 57 66 61
4 54 64 54
5 62 74 56
Sample Mean 55 68 57
Sample Variance 26.0 26.5 24.5
2008 Thomson South-Western. All Rights Reserved Slide 27
Testing for the Equality of k Population Means:
An Observational Study
p -Value and Critical Value Approaches
Develop the hypotheses.
H0: 1 = 2 = 3
Ha: Not all the means are equal
where:
1 = mean number of hours worked per
week by the managers at Plant 1
2 = mean number of hours worked per
week by the managers at Plant 2
3 = mean number of hours worked per
week by the managers at Plant 3
2008 Thomson South-Western. All Rights Reserved Slide 28
Testing for the Equality of k Population Means:
An Observational Study
ANOVA Table
Source of Sum of Degrees of Mean
Variation Squares Freedom Square F p-Value
Treatment 490 2 245 9.55 .0033
Error 308 12 25.667
Total 798 14
The p-value < .05, so we reject H0.
We have sufficient evidence to conclude that the
mean number of hours worked per week by
department managers is not the same at all 3 plant.
2008 Thomson South-Western. All Rights Reserved Slide 29
Multiple Comparison Procedures
Suppose that analysis of variance has provided
statistical evidence to reject the null hypothesis of
equal population means.
Fishers least significant difference (LSD) procedure
can be used to determine where the differences
occur.
2008 Thomson South-Western. All Rights Reserved Slide 30
Fishers LSD Procedure
Hypotheses
H 0 : i j
H a : i j
Test Statistic
xi x j
t
MSE( 1 n 1 n )
i j
2008 Thomson South-Western. All Rights Reserved Slide 31
Fishers LSD Procedure
Rejection Rule
p-value Approach:
Reject H0 if p-value <
Critical Value Approach:
Reject H0 if t < -ta/2 or t > ta/2
where the value of ta/2 is based on a
t distribution with nT - k degrees of freedom.
2008 Thomson South-Western. All Rights Reserved Slide 32
Fishers LSD Procedure
Based on the Test Statistic xi - xj
Example: Reed Manufacturing
Recall that Janet Reed wants to know
if there is any significant difference in
the mean number of hours worked per
week for the department managers
at her three manufacturing plants.
Analysis of variance has provided
statistical evidence to reject the null
hypothesis of equal population means.
Fishers least significant difference (LSD) procedure
can be used to determine where the differences occur.
2008 Thomson South-Western. All Rights Reserved Slide 33
Fishers LSD Procedure
Based on the Test Statistic xi - xj
LSD for Plants 1 and 2
Hypotheses (A) H 0 : 1 2
H a : 1 2
Rejection Rule
Reject H0 if x1 x2 > 6.98
Test Statistic
x1 x2 = |55 68| = 13
Conclusion
The mean number of hours worked at Plant 1 is
not equal to the mean number worked at Plant 2.
2008 Thomson South-Western. All Rights Reserved Slide 34
Fishers LSD Procedure
Based on the Test Statistic xi - xj
LSD for Plants 1 and 3
Hypotheses (B) H 0 : 1 3
H a : 1 3
Rejection Rule
Reject H0 if x1 x3 > 6.98
Test Statistic
x1 x3 = |55 57| = 2
Conclusion
There is no significant difference between the mean
number of hours worked at Plant 1 and the mean
number of hours worked at Plant 3.
2008 Thomson South-Western. All Rights Reserved Slide 35
Fishers LSD Procedure
Based on the Test Statistic xi - xj
LSD for Plants 2 and 3
Hypotheses (C) H 0 : 2 3
H a : 2 3
Rejection Rule
Reject H0 if x2 x3 > 6.98
Test Statistic
x2 x3 = |68 57| = 11
Conclusion
The mean number of hours worked at Plant 2 is
not equal to the mean number worked at Plant 3.
2008 Thomson South-Western. All Rights Reserved Slide 36
Type I Error Rates
The comparison-wise Type I error rate indicates
the level of significance associated with a single
pairwise comparison.
The experiment-wise Type I error rate EW is the
probability of making a Type I error on at least one of
the (k 1)! pairwise comparisons.
EW = 1 (1 )(k 1)!
The experiment-wise Type I error rate gets larger for
problems with more populations (larger k).
2008 Thomson South-Western. All Rights Reserved Slide 37
Chapter 13, Part B
Experimental Design and Analysis of Variance
Randomized Block Design
Factorial Experiments
2008 Thomson South-Western. All Rights Reserved Slide 38
Randomized Block Design
Experimental units are the objects of interest in the
experiment.
A completely randomized design is an experimental
design in which the treatments are randomly
assigned to the experimental units.
If the experimental units are heterogeneous, blocking
can be used to form homogeneous groups, resulting
in a randomized block design (single sample, each
block(element) is used in all treatments).
2008 Thomson South-Western. All Rights Reserved Slide 39
Randomized Block Design
ANOVA Procedure
For a randomized block design the sum of squares
total (SST) is partitioned into three groups: sum of
squares due to treatments, sum of squares due to
blocks, and sum of squares due to error.
SST = SSTR + SSBL + SSE
The total degrees of freedom, nT - 1, are partitioned
such that k - 1 degrees of freedom go to treatments,
b - 1 go to blocks, and (k - 1)(b - 1) go to the error term.
2008 Thomson South-Western. All Rights Reserved Slide 40
Randomized Block Design
ANOVA Table
Source of Sum of Degrees of Mean p-
Variation Squares Freedom Square F Value
SSTR MSTR
Treatments SSTR k-1 MSTR
k-1 MSE
Blocks SSBL b-1 SSBL
MSBL
b -1
SSE
Error SSE (k 1)(b 1) MSE
( k 1)(b 1)
Total SST nT - 1
2008 Thomson South-Western. All Rights Reserved Slide 41
Randomized Block Design
Example: Crescent Oil Co.
Crescent Oil has developed three
new blends of gasoline and must
decide which blend or blends to
produce and distribute. A study
of the miles per gallon ratings of the
three blends is being conducted to determine if the
mean ratings are the same for the three blends.
2008 Thomson South-Western. All Rights Reserved Slide 42
Randomized Block Design
Example: Crescent Oil Co.
Five automobiles have been
tested using each of the three
gasoline blends and the miles
per gallon ratings are shown on
the next slide.
Factor . . . Gasoline blend
Treatments . . . Blend X, Blend Y, Blend Z
Blocks . . . Automobiles
Response variable . . . Miles per gallon
2008 Thomson South-Western. All Rights Reserved Slide 43
Randomized Block Design
Type of Gasoline (Treatment)
Automobile Block
(Block) Blend X Blend Y Blend Z Means
1 31 30 30 30.333
2 30 29 29 29.333
3 29 29 28 28.667
4 33 31 29 31.000
5 26 25 26 25.667
Treatment
Means 29.8 28.8 28.4
2008 Thomson South-Western. All Rights Reserved Slide 44
Randomized Block Design
ANOVA Table
Source of Sum of Degrees of Mean
Variation Squares Freedom Square F p-Value
Treatments 5.20 2 2.60 3.82 .07
Blocks 51.33 4 12.80
Error 5.47 8 .68
Total 62.00 14
2008 Thomson South-Western. All Rights Reserved Slide 45
Factorial Experiments
In some experiments we want to draw conclusions
about more than one variable or factor.
Factorial experiments and their corresponding
ANOVA computations are valuable designs when
simultaneous conclusions about two or more factors
are required.
The term factorial is used because the experimental
conditions include all possible combinations of the
factors.
For example, for a levels of factor A and b levels of
factor B, the experiment will involve collecting data
on ab treatment combinations.
2008 Thomson South-Western. All Rights Reserved Slide 46
Chapter 13
Experimental Design and Analysis of Variance
Introduction to Experimental Design
and Analysis of Variance
Analysis of Variance
and the Completely Randomized Design
Multiple Comparison Procedures
2008 Thomson South-Western. All Rights Reserved Slide 47