Chapter 11
Module 2For self-study: Running Tests in SPSS
Assignment 2.ANOVA: Univariate
Evolution theory generates some hypotheses about the attractiveness of people’s faces. For instance, men with a strong chin and women with large eyes are assumed to be more appealing. A group of students was interested to see whether such hypotheses would apply to computer-generated characters appearing in games. To test their hypothesis they selected three male and three female faces and manipulated them using software they found on the internet. Their manipulations were aimed at making one set with faces that were assumed to be more appealing according to researchers in evolutionary aesthetics, and one set that was assumed to be less appealing according to the same criteria.
Open the file called evolution.sav to see whether the manipulation affected attractiveness. For this you run a Univariate analysis on all six pictures. In your analysis “condition” is the independent variable. Of course you need to take into account that men and women in the student sample may have different perceptions of what is attractive.
-
What would be a reason to use gender as either an independent variable or a covariate?
-
For this assignment use gender as a second independent variable in your analysis. Run Univariate analyses for each of the six pictures and interpret the results.
As to the choice between gender as a covariate or an independent variable (factor in the SPSS dialogue box) the students who conducted this study need to decide whether gender is part of their conceptual model or not. Are gender differences an essential element in the theories they examine? Did the theories generate hypotheses about such differences, and do they want to test them? In case they did, the students need to use gender as a second independent variable, next to condition. In case they do assume that gender plays a role, but are not primarily interested in the differences that may occur, or in the interaction that they might find between gender and the manipulations of male and female faces, they simply enter gender under “covariates” in the SPSS dialogue box. For the example it seems reasonable to assume that gender is a central variable in evolution theory and therefore also forms parts of the hypotheses of this study.
First, you to go to ANALYZE, then GENERAL LINEAR MODEL, then select UNIVARIATE from the available tests. You enter the variable for ratings on Picture 1 under dependent variable. Next, you transport gender and condition to the list of fixed factors.
Next you need to click on OPTIONS to make sure that that SPSS provides the descriptive statistics, that you also run a test of homogeneity, and, moreover, that you obtain a post hoc test.
Clicking on CONTINUE will bring you back to the main dialogue box. You then click on OK and SPSS will run your tests.
This is what you should see on your screen:
| Between-Subjects Factors | |||
|---|---|---|---|
| Value Label | N | ||
| Condition (Which version of the pictures was shown?) | 1 | Neutral | 20 |
| 2 | Evolutionary Unattractive | 21 | |
| 3 | Evolutionary Attractive | 19 | |
| Gender | 1 | Male | 29 |
| 2 | Female | 31 | |
This table gives you the distribution of the participants across conditions. Also, you see how many men and women participated. This is important information for the report students will have to write up.
| Descriptive Statistics | ||||
|---|---|---|---|---|
| Dependent Variable:Picture 1. How attractive do you think this person is? | ||||
| Condition (Which version of the pictures was shown?) | Gender | Mean | Std. Deviation | N |
| Neutral | Male | 5.44 | 2.455 | 9 |
| Female | 5.18 | 1.991 | 11 | |
| Total | 5.30 | 2.155 | 20 | |
| Evolutionary Unattractive | Male | 4.89 | 1.900 | 9 |
| Female | 5.25 | 2.301 | 12 | |
| Total | 5.10 | 2.095 | 21 | |
| Evolutionary Attractive | Male | 5.45 | 2.115 | 11 |
| Female | 5.87 | 1.356 | 8 | |
| Total | 5.63 | 1.802 | 19 | |
| Total | Male | 5.28 | 2.103 | 29 |
| Female | 5.39 | 1.944 | 31 | |
| Total | 5.33 | 2.006 | 60 | |
Now the second table already gives you an impression of what happened. On the ten-point scale the students responded, all scores are around 5, with little variation. For both women and men the manipulation of the female face in Picture 1 made hardly any difference. We will see in a minute whether the rest of the output confirms this. First however, look at the test of homogeneity that you ran.
Levene’s Test of Equality of Error Variances a| Dependent Variable:Picture 1. How attractive do you think this person is? | |||
|---|---|---|---|
| F | df1 | df2 | Sig. |
| .935 | 5 | 54 | .466 |
Tests the null hypothesis that the error variance of the dependent variable is equal across groups.
Levene’s Test of Equality of Error Variances a| Dependent Variable:Picture 1. How attractive do you think this person is? | |||
|---|---|---|---|
| F | df1 | df2 | Sig. |
| .935 | 5 | 54 | .466 |
Tests the null hypothesis that the error variance of the dependent variable is equal across groups.
a.Design: Intercept + GROUP + gender + GROUP * gender.
As you have read in Chapter 11, one of the conditions for running this test is equality of variance. For this purpose you ran Levene’s Test. As you can see, though, you have nothing to worry about, because the test is not significant, meaning that homogeneity of variance may be assumed. Now we look at the next table to study the results of the ANOVA.
Tests of Between-Subjects Effects| Dependent Variable:Picture 1. How attractive do you think this person is? | |||||
|---|---|---|---|---|---|
| Source | Type III Sum of Squares | df | Mean Square | F | Sig. |
| Corrected Model | 4.734 a | 5 | .947 | .220 | .953 |
| Intercept | 1682.093 | 1 | 1682.093 | 390.512 | .000 |
| GROUP | 3.474 | 2 | 1.737 | .403 | .670 |
| gender | .440 | 1 | .440 | .102 | .751 |
| GROUP * gender | 1.405 | 2 | .702 | .163 | .850 |
| Error | 232.600 | 54 | 4.307 | ||
| Total | 1944.000 | 60 | |||
| Corrected Total | 237.333 | 59 | |||
R Squared = .020 (Adjusted R Squared = −.071).
What do we see? The effect of condition (GROUP) was not significant (p = .670. This means that the manipulation did not affect perceived attractiveness. Second, you see that the factor GENDER had no effect either (p = .751). This means that the scores of men and women do not differ significantly. Also, as you can see, there is no interaction between the two variables (p = .850). This means that women and men did not respond differently to the manipulation of the faces. In conclusion, the students need to reject their hypotheses, or, in other words, to accept the null-hypothesis.