Statistical Methods for Psychology

Procedure.(Menu selections are not shown, but they should be evident.) This is the proce-

dure we will use in Chapters 13 and 14 and it produces the same answers as the one-way

procedure. It also produces some output that will not be familiar to you, which is explained

below.

Notice in the summary table that the first line is labeled “Corrected Model.” If there

were two or more independent variables (e.g., Group and Sex), then this line would repre-

sent the combined effects of those variables. Because there is only one independent vari-

able, the Model and the Group effects will be exactly the same.

The line labeled “Intercept” refers to a test on the null hypothesis that the grand mean

is equal to 0 in the population. We very rarely care about this test, although in this case it is

a test of a meaningful question about whether the girls in this study, averaged across

groups, gained weight.

The lines labeled “Group,” “Error,” and “Corrected total” are the same the results we

saw in Exhibit 11.1.

Notice that the entry of partial eta squared for the Group effect is the same as the

“R-squared” given at the bottom of the table. This tells us that 14% of the variation in

weight gain could be attributable to differences between treatments.

SPSS then calculates observed power, treating the obtained means as parameters,

and the obtained MSerroras an accurate estimate of the population variance. Because

there are unequal sample sizes in this example, you will have difficulty reproducing

these values exactly.

Section 11.13 Computer Analyses 355

Mean

−.450000

3.006897

7.264706

2.763889

Treatment Group

Control

CogBehav

Family

Total

Std.

Deviation

7.988705

7.308504

7.157421

7.983598

N

26

29

17

72

Descriptive Statistics

Dependent Variable: WTGAIN

Type III

Sum of

Squares

614.644b

732.075

614.644

3910.742

5075.400

4525.386

Source

Corrected Model

Intercept

GROUP

Error

Total

Corrected

Total

df

2

1

2

69

72

71

Mean

Square

307.322

732.075

307.322

56.677

F

5.422

12.917

5.422

Sig.

.006

.001

.006

Eta

Squared

.136

.158

.136

Noncent.

Parameter

10.845

12.917

10.845

Observed

Powera

.830

.943

.830

Tests of Between-Subjects Effects

Dependent Variable: WTGAIN

b R Squared = .136 (Adjusted R Squared = .111)

a Computed using alpha =.05

Exhibit 11.3 SPSS general linear model analysis of Everitt’s data on treatment of anorexia