Chapter 11
Module 2For self-study: Running Tests in SPSS
Assignment 4.Repeated Measures
In this assignment we will look at the results of a study by a group of students who were interested in the mere exposure hypothesis: by frequent exposure to a stimulus we get used to that stimulus to such an extent that we mix up familiarity with liking. In their study they also considered the opposite hypothesis, namely that after a great deal of exposure to one and the same stimulus people will get bored and may even reject the stimulus. To test this hypothesis, the students looked at the appreciation of the color orange, symbol of, among other things, the Dutch soccer team. During the World Cup, the entire country, the stores, the streets, the television screen is filled with orange objects. Some complain that after a while you cannot stand the color orange any more. Because it is so hard to ignore, the students decided to ask some friends of theirs who study outside the country to participate in their experiment as a control group. They had 10 participants in The Netherlands, and 9 abroad respond to five sets of four similar pictures. In every set there was one picture in orange, while the others were, for instance, yellow, purple and green. The participants were asked to divide 100 points over each of the four pictures in each set. The pretest was conducted just before the games started, the posttest just before The Netherlands were out of the contest. Both pretest and posttest consisted of five trials.
Open the file Orange.sav. First you need to calculate the average number of points attributed to the orange pictures. Check the information on this procedure in Chapter 7, section 5. In the variable view you will find the variables that pertain to the orange pictures. For the pretest that is “pre1.2”, “pre2.4”, “pre3.3”, “pre4.1”, “pre5.4”. In the posttest you will find the orange pictures are “post1.1”, “post2.3”, “post3.2”, “post4.4”, “post5.2”. Make sure that you have an average ‘pretest’ and a ‘posttest’ score, interpretable on a scale from 0–100. Call the pretest result “pretest”, and the posttest results “posttest”.
Second, run a Repeated Measures MANOVA to test whether “condition” (being in the Netherlands or abroad) affected scores, using “pretest” and “posttest” scores to register any possible changes in appreciation for the color orange. Interpret the results in the light of the two hypotheses that students were testing.
For this assignment you first have to compute two new variables: ‘pretest’ (average appreciation scores for the orange pictures shown right before the soccer frenzy), and ‘posttest’ (the same, but then some months later). To calculate the pretest results you probably did the following. Under the TRANSFORM menu you found COMPUTE. To obtain the correct outcome your dialogue box should look as follows:
And for the posttest, you should have entered the fields as follows:
Now you are ready to run the Repeated Measures MANOVA. For this you open the ANALYZE menu and look for the GENERAL LINEAR MODEL submenu. From that menu you select REPEATED MEASURES. If all went well, you entered the dialogue box as follows:
Click on ADD and then on DEFINE. The dialogue box that opens should be filled out as follows:
Click on OK and SPSS produces the test for you.
| Descriptive Statistics | ||||
|---|---|---|---|---|
| Condition | Mean | Std. Deviation | N | |
| Pretest | Experimental (exposure to orange) | 28.1600 | 6.95305 | 10 |
| Control (abroad during soccer frenzy) | 20.8889 | 10.98231 | 9 | |
| Total | 24.7158 | 9.57550 | 19 | |
| Posttest | Experimental (exposure to orange) | 23.1000 | 5.13052 | 10 |
| Control (abroad during soccer frenzy) | 26.8889 | 4.75511 | 9 | |
| Total | 24.8947 | 5.19503 | 19 | |
Looking at the table for descriptive statistics (you had to tick ‘descriptive statistics’ under OPTION S) you will see that the experimental group scored much higher on the pretest than the control group did. After a couple of months we see the pattern almost reversed: now the control group scores a little higher than the experimental group, and the scores of the experimental group dropped five points.
Multivariate Tests b| Effect | Value | F | Hypothesis df | Error df | Sig. | |
|---|---|---|---|---|---|---|
| orange | Pillai’s Trace | .002 | .037 a | 1.000 | 17.000 | .850 |
| Wilks’ Lambda | .998 | .037 a | 1.000 | 17.000 | .850 | |
| Hotelling’s Trace | .002 | .037 a | 1.000 | 17.000 | .850 | |
| Roy’s Largest Root | .002 | .037 a | 1.000 | 17.000 | .850 | |
| orange * GROUP | Pillai’s Trace | .232 | 5.125 a | 1.000 | 17.000 | .037 |
| Wilks’ Lambda | .768 | 5.125 a | 1.000 | 17.000 | .037 | |
| Hotelling’s Trace | .301 | 5.125 a | 1.000 | 17.000 | .037 | |
| Roy’s Largest Root | .301 | 5.125 a | 1.000 | 17.000 | .037 | |
Exact statistic.
b.Design: Intercept + GROUP.
Within Subjects Design: orange
Look now at the table above. What you see is a significant interaction for Orange (the repeated test for the appreciation of the color orange) and Group (participants living in The Netherlands and those participants living abroad).
The data of this study do not, however, seem to meet the criteria for a Multivariate analysis. The Maunchy’s test for Sphericity could not be computed. The students should therefore probably refer to a much simpler test: the paired samples t-test; but for the sake of illustration, let us see how the table below should be interpreted. You see that all the tests for within-subject effects reveal a significant interaction. This means that participants in the two conditions responded significantly different to the pretest as compared to the posttest. You can stop reading the output now. We have seen that in itself, the effect of the repeated measure (time) did not matter, but that we did see an interaction effect. In sum, we should conduct a simpler test, a t-test or a non-parametric equivalent before we can draw any conclusion.
Tests of Within-Subjects Effects| Measure:MEASURE_1 | ||||||
|---|---|---|---|---|---|---|
| Source | Type III Sum of Squares | df | Mean Square | F | Sig. | |
| orange | Sphericity Assumed | 2.093 | 1 | 2.093 | .037 | .850 |
| Greenhouse-Geisser | 2.093 | 1.000 | 2.093 | .037 | .850 | |
| Huynh-Feldt | 2.093 | 1.000 | 2.093 | .037 | .850 | |
| Lower-bound | 2.093 | 1.000 | 2.093 | .037 | .850 | |
| orange * GROUP | Sphericity Assumed | 289.714 | 1 | 289.714 | 5.125 | .037 |
| Greenhouse-Geisser | 289.714 | 1.000 | 289.714 | 5.125 | .037 | |
| Huynh-Feldt | 289.714 | 1.000 | 289.714 | 5.125 | .037 | |
| Lower-bound | 289.714 | 1.000 | 289.714 | 5.125 | .037 | |
| Error(orange) | Sphericity Assumed | 960.962 | 17 | 56.527 | ||
| Greenhouse-Geisser | 960.962 | 17.000 | 56.527 | |||
| Huynh-Feldt | 960.962 | 17.000 | 56.527 | |||
| Lower-bound | 960.962 | 17.000 | 56.527 | |||
The data of this study do not seem to meet the criteria for a Multivariate analysis. The Maunchy’s test for Sphericity could not be computed. The students should therefore probably refer to a much simpler test: the paired samples t-test; but for the sake of illustration, let us see how the table below should be interpreted. You see that all the tests for within-subject effects reveal a significant interaction. This means that participants in the two conditions responded significantly different to the pretest as compared to the posttest. You can stop reading the output now. We have seen that in itself, the effect of the repeated measure (time) did not matter, but that we did see an interaction effect. In sum, we should conduct a simpler test, a t-test or a non-parametric equivalent before we can draw any conclusion.