 
Mixed
Design ANOVAs
In some experiments the researcher makes use of two independent variables, one of which involves different groups of participants (a between subjects procedure) while all the levels of the other variable are given to all participants (within subjects design). The analysis of such complex designs is described here using an example from the verbal learning research. A psychologist interested in verbal learning wished to determine the effects of meaningfulness of material on the rate of learning. She randomly assigned participants to one of three groups, one of which must learn a low-meaningful list of nonsense syllables, the second a medium-meaningful list, and the third a high-meaningful list. All participants were given 15 trials, and the number of correct responses per trial was recorded. To simplify her analysis, she blocked the number of correct responses into three blocks of five trials each. Thus, the scores used in the analysis represent the number of correct responses for each successive block of five trials. The entry of data in this type of design combines that for within and between subjects procedures. The levels of the between-subject variable, meaning, is represented as 1, 2, and 3. While the within-subject variable (trial block) is presented in three separate columns, one for each block of trials. Also note the format for the variables id and meaning.
We begin our analysis by selecting Analyze, General Linear Model and finally, GLM -Repeated Measures.
Our first task is to define the within-subject factor (also know as Repeated Measures Factor). Call the factor trials and indicate that it has 3 levels. Then click the Add button. That will bring up this window.
Indicate the variables that represent the three levels of the within-subject IV by highlighting each and then pressing the arrow button. Your panel should look like this.
Now for the between-subject factor. Highlight meaning and insert in into the Between-Subjects Factor(s) window. This is what you should see on your screen.
Once the variables have been defined, we can begin selecting the options we want. Begin with Contrasts. Since we want contrasts for our within-subject variable only, select trails and then the Repeated option. Be sure to click Change to complete the task. Leave the between-subjects variable, meaning, with the None option selected. Press Continue to return to the main panel.
Now choose Plots. This is a complex design so we want lots of plots, including plots of each main effect as well as both interaction plots (trials as a function of meaning and meaning as a function of trials). To get a plot of the main effects, select a variable (meaning, for example) and then click the arrow to add it to the Horizontal Axis box. The click the Add option near the bottom of the screen. Do this for both meaning and trials.
To get an interaction of meaning by trials, with levels of trials on the horizontal axis, begin just as you would to make a plot of the main effect for trials. But before you 'add' the plot, highlight meaning and use the arrow button to put it in the Separate Lines box. Then 'add' it to the plot list. Follow the same procedure to get an interactive plot with levels of meaning on the horizontal axis. Your final panel should look like the one above. (Press Continue to return to the main panel.) PostHoc options are next. Highlight the between-subject variable (meaning) and move it to the PostHoc Tests for window. The Tukey test is a reasonable selection here to check that box. (Continue…)
Our last set of choices involves Options. (See panel below.) From the Factor(s) and Factor Interactions list, select meaning, trials, and meaning*trials and move them to the Display Means for box. Then select descriptive statistics, effect sizes, and homogeneity tests. Your window should then look like the one below. Continue to return to main panel.
To begin the analysis, press OK on the main panel.
5/00
|
