SSRIC Teaching Resources Depository
PERS Module
Larry Herringer, Psychology Deptartment
California State University, Chico

Chapter 9: Comparing Mean Levels of Behavior

This exercise illustrates the creation of comparison groups from a continuous measure, and 1-Way Analysis of Variance. It uses the PERS dataset, consisting of 90 cases and 968 variables. The variables represent measures of traits and relevant behaviors for the dimensions of extraversion (outgoingness) and conscientiousness, reported each week for three weeks by a group of undergraduate psychology students.

If we grouped participants into "low," "average," and "high" outgoingness groups, based on their personality trait scores, would we find that these groups of people are significantly different in their behavior?

Instead of correlating continuous measures with each other to examine the relationships between variables, many researchers in psychology prefer to measure behaviors or other responses, and then compare the mean levels of these across different experimental conditions or participant groups. Analysis of Variance is a very useful and powerful technique for this, and not surprisingly it is widely used throughout the behavioral sciences.

This exercise will use 1-Way ANOVA to compare behaviors across levels of outgoingness, as measured by the JPI Social Participation scale (jpispt). First, obtain a frequency distribution of jpispt scores for our participants by using:

Analyze>Descriptive Statistics>Frequencies

The output shows that our sample has jpi scores ranging from 1 to 20 (the full range of the scale). According to the test manual (see the reference in the codebook), the mean is about 11 and the standard deviation is about 4, for a college population. If we use the heuristic that most personality traits are distributed quasi-normally, so that most people are "average" in the trait, we can divide our participants into high, average, and low outgoingness. One standard deviation from the mean in either direction would demarcate a middle "average" group from those at the extreme high or low. To keep things simple, we will use 1-6 for low, 7-15 for average, and 8-20 to define high outgoingness.

We need to use an "If" command with the Transform>Compute menu; since we will be doing this 3 times, it is easier to type commands directly by opening a syntax window:

File>New>Syntax.

Type the following in the syntax window:

If (jpispt<7) extgrp=1.

If (jpispt>=7 and jpispt<=15) extgrp=2.

If (jpispt>15) extgrp=3.

In the syntax window menu, type on Run>All.

This creates a new variable, "extgrp" which is coded "1" for low outgoingness participants, "2" for those who are average, and "3" for those who are high. We can now use this variable to compare the three groups on any behavior we choose. You can add labels if you want (e.g., 1="low") by clicking on the column heading ("extgrp" in gray) for this variable in the SPSS Data Editor window.

Use the menu Analyze>Compare Means>One-Way ANOVA

Dependent List: wbo1

Factor: extgrp

If your results indicate that the three groups are significantly different, you can re-run the analysis, this time selecting Options such as Descriptives and Post-hocs such as the Tukey test, to compare the groups individually. You should find that this analysis does NOT indicate that the three groups are significantly different on wbo1 (number of days the participant ate breakfast with other people). Try using an aggregate instead, such as the 2-item (wbo7 & wbo27) aggregate noted in Exercise 8. You should find strong differences between the three groups for that measure. Try different behaviors, or even different traits, for group comparisons. What is the difference between this approach to showing how personality is related to behavior, and the correlation approach of the previous exercises. Are there advantages to each?