AMS 572 Lecture Notes

¨ ¥
§AMS 572 Lecture Notes ¦
Chapter 13. One-Way Analysis of Variance (ANOVA)

• This is an extension of the pooled-variance t-test.
• The pooled-variance t-test tests the null hypothesis that two population
means are equal, i.e. H0 : µ1 = µ2 .
• The One-way ANOVA can test the equality of several population means.
That is:
H0 : µ1 = µ2 = µ3 = · · · = µt
Ha : At least one of the t-population means differs from the rest.
• Assumptions
1. Normal populations.
2. equality of population variances, σ12 = σ22 = · · · = σt2 .
• Test statistic:
s2B H0
F = ∼ Ft−1,nT −t
s2W
SSB SSW
where s2B = , s2W = .
t−1 nT − t
Xt
2
Here SSB = ni (ȳi· − ȳ·· ) (Sum of squares between samples);
i=1
Xt Xni
SSW = (yij − ȳi· )2 (Sum of squares within samples)
i=1 j=1
= T T S − SSB ;
Xt X ni
T SS = (yij − ȳi· )2 (Total sum of squares);
i=1 j=1
1
X y··2
2
T SS = yij −
i,j
nT
X y2 y2
SSB = i i· − ··
ni nT
SSW = T SS − SSB.
yij : The jth sample observation selected from population i. For example,
y23 denotes the third sample observation drawn from population 2.
ni : The number of sample observations selected from population i. In our
data set, n1 , the number of observations obtained from population
1, is 4. Similarly, n2 = n3 = n4 = n5 = 4. However, it should be
noted that the sample sizes need not be the same. Thus, we might
have n1 = 12, n2 = 3, n3 = 6, n4 = 10, and so forth.
P
nT : The total sample size; nT = ni . For the data given in Table 13.4,
nt = n1 + n2 + n3 + n4 + n5 = 20.
yi : The sum (total) of the sample measurements obtained from popula-
tion i.
P
y·· : The sum (grand total) of all sample observations: y·· = yi· .
ȳi· : The average of the ni sample observations drawn from population i,
ȳi· = yi /ni .
ȳ·· : The average of all sample observations; ȳ·· = y·· /nT .
α
• R.R. : We reject H0 in favor of Ha if the test statistic F exceeds FU,t−1,n T −t
.
This test is often summarized in an analysis of variance table. (AOV ta-
ble).
Sum of Degrees of Mean

Source Squares Freedom Square F Test
Between samples SSB t−1 s2B = SSB/(t − 1) s2B /s2W
2
Within samples SSW nT − t sW = SSW/(nT − t)
Totals TTS nT − 1
¤ ¡
£EXAMPLE 8.1 ¢
A clinical psychologist wished to compare three methods for reducing hos-
tility levels in university students. A certain test (HLT) was used to mea-
sure the degree of hostility. A high score on the test indicated great hos-
tility. Eleven students obtaining high and nearly equal scores were used in
the experiment. Four were selected at random from among the 11 problem
2
cases and treated with method 1. Four of the remaining seven students
were selected at random and treated with method 2. The remaining three
students were treated with method 3. All treatments were continued for
a one-semester period. Each student was given the HLT test at the end
of the semester, with the results shown in Table 13.8. Use these data to
perform an analysis of variance to determine if there are differences among
mean scores for the three methods. Use α = .05.
Method Test Scores Totals

1 80 92 87 83 342
2 70 81 78 74 303
3 63 76 70 209
Total 854
Solution: The null and alternative hypotheses are
H0 : µ1 = µ2 = µ3
Ha : At least one of the population means differes from the rest.
For n1 = 4, n2 = 4, and n3 = 3, we have a total sample size of nT = 11.

The totals from Table are
y1· = 342, y2· = 303, y3· = 209, y·· = 854.
Substituting into the computational formulas for TSS and SSB, we have
X y··2 (854)2
2
T SS = yij − = (80)2 + (92)2 + · · · + (70)2 −
i,j
nT 11
= 66, 988 − 66, 301.45 = 686.55
X y2 y2 (342)2 (303)2 (209)2
SSB = i·
− ·· = + + − 66, 301.45
ni nT 4 4 3
= 66, 753.58 − 66, 301.45 = 452.13.
Then
SSW = 686.55 − 452.13 = 234.42.
The AOV table for these data is shown in Table 13.9.
3
Source SS df MS F
Between samples 452.13 2 226.07 226.07/29.3=7.72
Within samples 234.42 8 29.30
Totals 686.55 10
The critical value of F is obtained from Table 6 in the Appendix for α = .5,
df 1 = 2, and df 2 = 8; this value is 4.46. Since the computed value of F,
7.72, exceeds the tabulated value, 4.46, we reject the null hypothesis of
equality of the mean scores for the three groups. Computer output shown
here verifies the results we obtained by hand.
LISTING OF DATA
OBS METHOD SCORE
1 1 80
2 1 92
3 1 87
4 1 83
5 2 70
6 2 81
7 2 78
8 2 74
9 3 63
10 3 76
11 3 70
ANALYSIS OF VARIANCE PROCEDURE

GENERAL LINEAR MODELS PROCEDURE
DEPENDENT VARIABLE: SCORE
SOURCE DF SUM OF SQUARES MEAN SQUARE F VALUE PR > F R-SQUARE C.V
MODEL 2 452.12878788 226.06439394 7.71 0.0136 0.658556 6.9724
ERROR 8 234.41666667 29.30208333 ROOT MSE SCORE MEAN
CORRECTED TOTAL 10 686.54545455 5.41313988 77.63636364
SOURCE DF TYPE I SS F VALUE PR > F DF TYPE III SS F VALUE PR > F

METHOD 2 452.12878788 7.71 0.0136 2 452.12878788

AMS 572 Lecture Notes

Uploaded by

Copyright:

Available Formats

AMS 572 Lecture Notes

Uploaded by

Document Information

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

AMS 572 Lecture Notes

Uploaded by

Copyright:

Available Formats

¨ ¥

§AMS 572 Lecture Notes ¦

Chapter 13. One-Way Analysis of Variance (ANOVA)

Sum of Degrees of Mean

Method Test Scores Totals

Solution: The null and alternative hypotheses are

For n1 = 4, n2 = 4, and n3 = 3, we have a total sample size of nT = 11.

y1· = 342, y2· = 303, y3· = 209, y·· = 854.

The AOV table for these data is shown in Table 13.9.

ANALYSIS OF VARIANCE PROCEDURE

SOURCE DF TYPE I SS F VALUE PR > F DF TYPE III SS F VALUE PR > F

You might also like