10 Anova

9/3/2018
ANOVA
www.utm.my innovative ● entrepreneurial ● global
ANOVA – Independent samples
1
9/3/2018
The Logic and the Process of Analysis of Variance
• ANOVA can be used in situations where there are

two or more means being compared, whereas the t
tests are limited to only two means are involved.
• These situations would require a series of several t
tests to evaluate all of the mean differences. (t test
can compare only 2 means at a time.)
• Although each t test can be done with a specific α-
level (risk of Type I error), the α-levels accumulate
over a series of tests so that the final
experimentwise α-level can be quite large
• ANOVA allows researcher to evaluate all of the mean

differences in a single hypothesis test using a single
α-level and, thereby, keeps the risk of a Type I error
under control no matter how many different means
are being compared.
2
9/3/2018
• The test statistic for ANOVA is an F-ratio, which is a

ratio of two sample variances. In the context of
ANOVA, the sample variances are called mean
squares, or MS values.
• The top of the F-ratio MSbetween measures the size of
mean differences between samples. The bottom of
the ratio MSwithin measures the magnitude of
differences that would be expected without any
treatment effects.
3
9/3/2018
4
9/3/2018
The Logic and the Process of Analysis of Variance

(cont.)
Example
• One experiment call examining learning
performance under three temperature
condition. The dependent variable is the number
of problem solved correctly.
Temperature conditions Is there any
50o 70o 90o significant
0 4 1 different
between three
1 3 2
groups at alpha
3 6 2
0.05?
1 3 0
0 4 0
5
9/3/2018
Steps
1. State the hypothesis
Ho: µ1= µ2= µ3
H1: At least one of the treatment means is different.
2. Find the critical value with α = 0.05 and F(between, within)
4. Make decision
Example 1
Temperature conditions
50o 70o 90o
0 4 1
1 3 2
3 6 2
1 3 0
0 4 0
6
9/3/2018
Example
Temperature conditions
50o 70o 90o
0 4 1
1 3 2 G=30
3 6 2 N=15
1 3 0 k=3
0 4 0
T1=5 T2=20 T3=5
SS1=6 SS2=6 SS3=4
n1=5 n2=5 n3=5
Source SS df MS F
Between SSbetween dfbetween
subject subject
Within SSwithin dfwithin
treatment
Total SStotal dftotal
7
9/3/2018
SPSS OUTPUT
ANOVA
ANSWER
Sum of Mean
Squares df Square F Sig.
Between
Groups
Within Groups
Total
SPSS OUTPUT
ANOVA
ANSWER
Sum of Mean
Between
Groups 30.00 2 15.00 11.25 .002
Within Groups
16.00 12 1.33
Total
46.00 14
8
9/3/2018
Analysis of Variance and Post Tests
• The null hypothesis for ANOVA states that for the

general population there are no mean differences
among the treatments being compared; H0: μ1 = μ2 =
μ3 = . . .
• When the null hypothesis is rejected, the conclusion
is that there are significant mean differences.
• However, the ANOVA simply establishes that
differences exist, it does not indicate exactly which
treatments are different.
Analysis of Variance and Post Tests
• With more than two treatments, this creates a

problem. Specifically, you must follow the ANOVA
with additional tests, called post tests, to determine
exactly which treatments are different and which are
not.
• The Scheffe test and Tukey=s HSD are examples of
post tests.
• These tests are done after an ANOVA where H0 is
rejected with more than two treatment conditions.
The tests compare the treatments, two at a time, to
test the significance of the mean differences.
9
9/3/2018
POST -HOC
Multiple Comparisons
ANSWER
Tukey HSD
Mean 95% Confidence Interval
Difference (I- Lower Upper
(I) GROUP (J) GROUP J) Std. Error Sig. Bound Bound
50 70 -3.000* .730 .004 -4.95 -1.05
90 .000 .730 1.000 -1.95 1.95
70 50 3.000* .730 .004 1.05 4.95
90 3.000* .730 .004 1.05 4.95
90 50 .000 .730 1.000 -1.95 1.95
70 -3.000* .730 .004 -4.95 -1.05
*. The mean difference is significant at the 0.05 level.
Measuring Effect Size for an Analysis of Variance
• To supplement the hypothesis test, it is

recommended that you calculate a measure of effect
size.
• For an analysis of variance the common technique
for measuring effect size is to compute the
percentage of variance that is accounted for by the
treatment effects.
10
9/3/2018
• For the t statistics, this percentage was identified as

r2, but in the context of ANOVA the percentage is
identified as η2 (the Greek letter eta, squared).
• The formula for computing effect size is:
SSbetween treatments
η2 = ───────────
SStotal
30
η2 = ─────────── = 0.65
46
11
9/3/2018
Example 2
Treatment
A B C
2 3 7
6 7 5
2 6 4
6 4 8
Example 2
Treatment
A B C
2 3 7 G=60
6 7 5 N=12
2 6 4
k=3
6 4 8
T1=16 T2=20 T3=24
SS1=16 SS2=10 SS3=10
n1=4 n2=4 n3=4
12
9/3/2018
Source SS df MS F
subject subject
treatment
One Way Anova

Source SS df MS F
subject subject
treatment
13
9/3/2018
SPSS OUTPUT
ANOVA
ANSWER
Sum of Mean
Between
Groups
Within Groups
Total
SPSS OUTPUT
ANOVA
ANSWER
Sum of Mean
Between
Groups 8 2 4 1 0.405
Within Groups
36 9 4
Total
44 11
14
9/3/2018
ANOVA TABLE
ANSWER
Sum of Mean
Squares df Square F
Between
Groups
Within Groups
Total
15
9/3/2018
ANOVA TABLE
ANSWER
Sum of Mean
Squares df Square F
Between
Groups 468.7 2 234.35 28.53
Within Groups
345 42 8.21
Total
813.7 44
SPSS
ANOVA
Independent-measures
16
9/3/2018
Is there any significant differences

between Mathematics Dec score among 4
ethnic group?
17
9/3/2018
18
9/3/2018
Reject null hypothesis
19
9/3/2018
20
9/3/2018
SPSS- FOR ORDINAL DATA
SPSS
Non parametric
Independent-measured
Kruskal-Wallis Test
For ordinal data
Note
• Ratio or interval data can be ranked and can
consider as an ordinal data.
• When the data cannot satisfied the
requirement to run ANOVA, go for non
parametric.
21
9/3/2018
Is there any significant differences

between Mathematics Jan score among 4
ethnic groups?
SPSS
Data MJAN is not normally distributed.

Go for non parametric test
22
9/3/2018
23
9/3/2018
24
9/3/2018
ANOVA
Repeated-Measures
The Logical Background for a Repeated-

Measures ANOVA
• The repeated-measures design eliminates individual

differences from the between-treatments variability
because the same subjects are used in every
treatment condition.
• The result is a test statistic similar to the
independent-measures F-ratio but with all individual
differences removed.
25
9/3/2018
Comparing Independent-Measures and

Repeated-Measures ANOVA
• The independent-measures analysis is used in

research situations for which there is a separate
sample for each treatment condition.
• The analysis compares the mean square (MS)
between treatments to the mean square within
treatments in the form of a ratio
MS between treatments
F = ───────────
MS within treatments

• What makes the repeated-measures analysis different

from the independent-measures analysis is the
treatment of variability from individual differences.
The independent-measures F ratio (Chapter 13) has the
following structure:
MS between treatments treatment effect + error (including individual differences)
F = ──────────── = ───────────────────────────────────
MS within error (including individual differences)
• In this formula, when the treatment effect is zero (H0

true), the expected F ratio is one.
26
9/3/2018

• In the repeated-measures study, there are no

individual differences between treatments because
the same individuals are tested in every treatment.
• This means that variability due to individual
differences is not a component of the numerator of
the F ratio.
• Therefore, the individual differences must also be
removed from the denominator of the F ratio to
maintain a balanced ratio with an expected value of
1.00 when there is no treatment effect.

• That is, we want the repeated-measures F-

ratio to have the following structure:
treatment effect + error (without individual differences)

F = ────────────────────────────────────
error (with individual differences removed)
27
9/3/2018

• This is accomplished by a two-stage analysis. In the
first stage, total variability (SS total) is partitioned into
the between-treatments SS and within-treatments
SS.
• The components for between-treatments variability
are the treatment effect (if any) and error.
• Individual differences do not appear here because
the same sample of subjects serves in every
treatment. On the other hand, individual differences
do play a role in SS within because the sample contains
different subjects.

• In the second stage of the analysis, we

measure the individual differences by
computing the variability between subjects, or
SS between subjects
• This value is subtracted from SS within leaving a
remainder, variability due to experimental
error, SS error
28
9/3/2018
Example 1
Participant 24H 36H 48H

A 0 0 6
B 1 3 5
C 0 1 5
D 4 5 9
E 0 1 5
29
9/3/2018
Steps
1. State the hypothesis
Ho: µ1= µ2= µ3
H1: At least one of the treatment means is different.
2. Find the critical value with α = 0.05
4. Make decision
30
9/3/2018
Example 1
Participant 24H 36H 48H P Total

A 0 0 6 6 N=15
B 1 3 5 9 G=45
C 0 1 5 6
D 4 5 9 18
E 0 1 5 6
T 5 10 30
SS 12 16 12
31
9/3/2018
ANOVA table
Source SS df MS F

Between Subject Ssbetween subject dfbetween subject
Error SSerror dferror
ANOVA table
Source SS df MS F
Between
Within
Between Subject
Error
Total
32
9/3/2018
ANOVA table
Source SS df MS F
Between 70 2 35 70
Within 40 12
Between Subject 36 4
Error 4 8 0.5
Total 110 14
SPSS
Tests of Within-Subjects Effects
Measure:MEASURE_1
Type III Sum of

Source Squares df Mean Square F Sig.
factor1 Sphericity Assumed 70.000 2 35.000 70.000 .000
Greenhouse-Geisser 70.000 1.000 70.000 70.000 .001
Huynh-Feldt 70.000 1.000 70.000 70.000 .001
Lower-bound 70.000 1.000 70.000 70.000 .001
Error(factor1) Sphericity Assumed
4.000 8 .500
Greenhouse-Geisser
4.000 4.000 1.000
Huynh-Feldt
4.000 4.000 1.000
Lower-bound
4.000 4.000 1.000
33
9/3/2018
Example 2
Depression Scores
Sample Before Therapy After Therapy 6 Month
Follow up
A 71 53 55
B 62 45 44
C 82 56 61
D 77 50 46
E 81 54 55
Example 2
Depression Scores
Sample Before Therapy After Therapy 6 Month Person
Follow up Total
A 71 53 55 179 n=5
B 62 45 44 151 K=3
C 82 56 61 199 N=15
D 77 50 46 173 G=892
E 81 54 55 190
T=373 T=258 T=261
SS=273.2 SS=73.2 SS=198.8
34
9/3/2018
ANOVA table
Source SS df MS F

Between Subject Ssbetween subject dfbetween subject
Error SSerror dferror
ANOVA table
Source SS df MS
Between
Within
Between Subject
Error
Total
35
9/3/2018
ANOVA table
Source SS df MS
Between 1718.5 2 859.25 69.6
Within 545.2 12
Between Subject 446.4 4
Error 98.8 8 12.35
Total 2263.7 14
Measuring Effect Size for the Repeated-

Measures Analysis of Variance
• The formula for computing effect size for a

repeated-measures ANOVA is:
SS between treatments SS between treatments

η2 = ───────────── = ──────────────
SS total - SS between subjects SS error + SS between treatments
36
9/3/2018
SPSS
ANOVA
Repeated-Measures
Is there any significant differences in

Mathematics score for Jan, June and Dec?
37
9/3/2018
38
9/3/2018
39
9/3/2018
40
9/3/2018
SPSS-Ordinal data
SPSS
Non Parametric
Repeated-Measures
Friedman Test
Is there any significant differences on

perception of favourite teacher among
student for Jan, June and Dec?
41
9/3/2018
Data
42
9/3/2018
Ranks
Mean Rank
SUM OF FI FOR JAN 2.75
SUM OF FI FOR JUN 1.63
SUM OF FI FOR DEC 1.62
Test Statisticsa
N 42
Chi-Square 38.662
df 2
Asymp. Sig. .000
a. Friedman Test
43
9/3/2018
Two Factor ANOVA (Independent Measures)
• Also known as factorial design

• Involved two independent variables (factor A
and factor B) one dependent variable (M1,M2,M3,M4)
FACTOR A
YES NO
LOW M1 M3
FACTOR B
HIGH M2 M4
• M = mean
44
9/3/2018
FACTOR A
YES NO
LOW M1 M3
FACTOR B
HIGH M2 M4
• Hypothesis Null 1: No different mean between

group LOW and HIGH?
• Hypothesis Null 2: No different mean between
group YES and NO?
• Hypothesis Null 3: No interaction between factor
A & B.
• The mean differences between columns or

mean differences between rows simply
describe the main effects for two-factor study.
• Interaction between two factors occurs
whenever the mean differences between
individual treatment conditions, or cell, are
different from what would be predicted from
the overall main effects of the factors
45
9/3/2018
Factor A: Ethnic group

Factor B: Future study place
Dependent variable: Math Score
1. No mean different between ethnic group

2. No mean different between choices of future
study place.
3. No interaction between future study place and
ethnic group
46
9/3/2018
47
9/3/2018
48
9/3/2018
With alpha 0.05

All three hypothesis fail to reject.
49

10 Anova

Uploaded by

Copyright:

Available Formats

You might also like

10 Anova

Uploaded by

Document Information

Original Description:

Original Title

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

10 Anova

Uploaded by

Copyright:

Available Formats

9/3/2018

www.utm.my innovative ● entrepreneurial ● global

ANOVA – Independent samples

www.utm.my innovative ● entrepreneurial ● global

The Logic and the Process of Analysis of Variance

• ANOVA can be used in situations where there are

www.utm.my innovative ● entrepreneurial ● global

• ANOVA allows researcher to evaluate all of the mean

www.utm.my innovative ● entrepreneurial ● global

www.utm.my innovative ● entrepreneurial ● global

• The test statistic for ANOVA is an F-ratio, which is a

www.utm.my innovative ● entrepreneurial ● global

www.utm.my innovative ● entrepreneurial ● global

www.utm.my innovative ● entrepreneurial ● global

The Logic and the Process of Analysis of Variance

www.utm.my innovative ● entrepreneurial ● global

www.utm.my innovative ● entrepreneurial ● global

www.utm.my innovative ● entrepreneurial ● global

www.utm.my innovative ● entrepreneurial ● global

www.utm.my innovative ● entrepreneurial ● global

www.utm.my innovative ● entrepreneurial ● global

www.utm.my innovative ● entrepreneurial ● global

www.utm.my innovative ● entrepreneurial ● global

Analysis of Variance and Post Tests

• The null hypothesis for ANOVA states that for the

www.utm.my innovative ● entrepreneurial ● global

Analysis of Variance and Post Tests

• With more than two treatments, this creates a

www.utm.my innovative ● entrepreneurial ● global

www.utm.my innovative ● entrepreneurial ● global

Measuring Effect Size for an Analysis of Variance

• To supplement the hypothesis test, it is

www.utm.my innovative ● entrepreneurial ● global

Measuring Effect Size for an Analysis of Variance

• For the t statistics, this percentage was identified as

www.utm.my innovative ● entrepreneurial ● global

Measuring Effect Size for an Analysis of Variance

www.utm.my innovative ● entrepreneurial ● global

www.utm.my innovative ● entrepreneurial ● global

www.utm.my innovative ● entrepreneurial ● global

www.utm.my innovative ● entrepreneurial ● global

One Way Anova

www.utm.my innovative ● entrepreneurial ● global

www.utm.my innovative ● entrepreneurial ● global

www.utm.my innovative ● entrepreneurial ● global

www.utm.my innovative ● entrepreneurial ● global

www.utm.my innovative ● entrepreneurial ● global

www.utm.my innovative ● entrepreneurial ● global

www.utm.my innovative ● entrepreneurial ● global

Is there any significant differences

www.utm.my innovative ● entrepreneurial ● global

www.utm.my innovative ● entrepreneurial ● global

www.utm.my innovative ● entrepreneurial ● global

www.utm.my innovative ● entrepreneurial ● global

www.utm.my innovative ● entrepreneurial ● global

Reject null hypothesis

www.utm.my innovative ● entrepreneurial ● global