Introduction: What Is a Repeated-Measures Design?

A one-way repeated-measures ANOVA is appropriate when:

• the analysis involves a single predictor variable measured on a nominal scale;

• the analysis also involves a single criterion variable measured on an interval- or ratio-scale;

• each participant is exposed to each condition under the independent variable.

The repeated-measures design derives its name from the fact that each participant provides repeated scores on the criterion variable. Each participant is exposed to every treatment condition under the study’s independent variable and provides scores on the criterion under each of these conditions. Perhaps the easiest way to understand the repeated-measures design is to contrast it with the between-subjects design in which each participant participates in only one treatment condition.

For example, Chapter 9, “One-Way ANOVA with One Between-Subjects Factor,” presents a simple experiment that uses a between-subjects design. In that fictitious study, participants were randomly assigned to one of three experimental conditions. In each condition, they read descriptions of a number of fictitious romantic partners and rated their likely commitment to each partner. The purpose of that study was to determine whether the “level of rewards” associated with a given partner affected the participant’s rated commitment to that partner. Therefore, the level of rewards was manipulated by varying the description of one specific partner (partner 10) that was presented to the three groups. The level of rewards was manipulated in this way:

• Participants in the low-reward condition read that partner 10 provides few rewards in the relationship.

• Participants in the mixed-reward condition read that partner 10 provides mixed rewards in the relationship.

• Participants in the high-reward condition read that partner 10 provides many rewards in the relationship.

After reading this description, each participant rated how committed he or she would probably be to partner 10.

This study was called a between-subjects study because the participants were divided into different treatment groups, and the independent variable was manipulated between these groups. In a between-subjects design, each participant is exposed to only one level of the independent variable. (In the present case, that means that a given participant read either the low-reward description, the mixed-reward description, or the high-reward description, but no participant read more than one of these descriptions of partner 10.)

In a repeated-measures design on the other hand, each participant is exposed to every level of the independent variable and provides scores on the dependent variable under each of these levels. For example, you could easily modify the preceding study so that it becomes a one-factor repeated-measures design. Imagine that you conduct your study with a single group of 20 participants rather than with three treatment groups. You ask each participant to go through a list of potential romantic partners and rate his or her commitment to each partner. Imagine further that all three versions of partner 10 appear somewhere in this list, and that a given participant responds to each of these versions individually.

For example, a given participant might be working her way through the list and the third potential partner she comes to happens to be the low-reward version of partner 10 (assume that you have renamed this fictitious partner to be “partner 3”). She rates her commitment to this partner and moves on to the next partner. Later, the 11th partner she comes to happens to be the mixed-reward version of partner 10 (now renamed “partner 11”). She rates her commitment and moves on. Finally, the 19th partner she comes to happens to be the high-reward version of partner 10 (now renamed as “partner 19”). She rates her commitment and moves on.

Your study now uses a repeated-measures design because each participant has been exposed to all three levels of the independent variable. To analyze these data, you would create three SAS variables to include the commitment ratings made under the three different conditions:

• One variable (perhaps named LOW) contains the commitment ratings made for the low-reward version of the fictitious partner.

• One variable (perhaps named MIXED) contains the commitment ratings made for the mixed-reward version of the fictitious partner.

• One variable (perhaps named HIGH) contains the commitment ratings made for the high-reward version of the fictitious partner.

To analyze your data, you would compare scores for each of the three variables. Perhaps you would hypothesize that the commitment score contained in HIGH would be significantly higher than the commitment scores contained in LOW or MIXED.

Make note of two cautions before moving on. First, remember that you need a special type of statistical procedure to analyze data from a repeated-measures study such as this. You should not analyze these data using the program for a one-way ANOVA with one between-subjects factor as was illustrated in Chapter 9. This chapter shows you the appropriate SAS program for analyzing repeated-measures data.

Second, the fictitious study described here was used merely to illustrate the nature of a repeated-measures research design; do not view it as an example of a good repeated-measures research design. (In fact, the preceding study suffers from several serious problems.) This is because repeated-measures studies are vulnerable to a number of problems that are not encountered with between-subjects designs. This means that you must be particularly concerned about design when your study includes a repeated-measures factor. Some of the problems associated with this design are discussed later in the section “Sequence Effects.”