Lab Exercise--Repeated Measures  Analysis

of Variance #2

2/21/02

Announcements

I will be out of town for this lab. I want people to work through it carefully, so as to understand the general structure of the analysis.

More Complex Designs

I’m going to skip the repeated measures design with 2 Between-Subject measures, because I think it is pretty easy to understand.

One Between-Subj. Measure and Two Within-Subj Measures

Jenkins, Myerson, Joerding, and Hale (2000) published a paper on "Converging evidence that visuospatial cognition is more age-sensitive than verbal cognition" in Psychology of Aging, 15, 157-175. This is a very sophisticated journal from a statistical point of view, and studies are not always easy to understand. In fact, I had to work hard to understand this one, and then work even harder to make the data come out.

Jenkins et al. presented young and older participants with visual stimuli. They simply had to report whether there were 10 or 11 stimuli on the screen. The screen had a bunch of boxes (10 or 11) scattered over it, with a letter in the center of each box. For one condition (Visual) the boxes came in fixed patterns, and the participant had to learn that certain patters were associated with 10 stimuli, and others with 11 stimuli. The letter in the box was irrelevant. For the other condition (Verbal), the pattern of the boxes on the screen was irrelevant, and participants had to learn that certain letters went with 10 and other letters with 11 stimuli. these are the visual and verbal conditions, respectively.

Participants, as I said, were either young or older, and that is obviously a between subject variable. But all participants served under both the verbal and visual condition, so that is within-subjects. In addition, each participant sat through 6 blocks of trials (they used 9) for each condition, making Blocks another repeated measure. The authors had about 16 participants in each group, but I needed to use approximately 50 to make things come out. I also had to fudge the variances and play with the correlations. My conclusions are in line with theirs, but that's about it.

The dependent variable was a weirdly adjusted score, and you can read the paper if you want to find out what it really was. Consider it a measure of accuracy, where low scores are better.

The data are available on Gumby , and I will give someone a disk. Alternatively, you can download them as Jenkins.sav. For the Age variable, 1 is Young and 2 is Older. The dependent variables are named pverbal1 -- pvisual6, corresponding the the Task  (Verbal or Visual), and the Block (1 through 6). 

Call up the data and work through getting the cell means. This will require thinking about what order the means should be in.

Next,  plot the means of the important interactions by hand

For some plots it is far easier to plot by hand than by SPSS . (You do not have to be precise.)

Notice that we can't squeeze them all on one plot--we need separate plots either for the Modality or for Age.

What do the data suggest?

Set this up as a repeated measures ANOVA. 

You probably have to work together and help each other on this.

•     Tell SPSS that you want a repeated measures design.
•     Label the first within subjects factor as Modality with 2 levels (verbal and visual).
•     Label the second within subjects factor as Trials with 6 levels.
•     Click "define" and assign pverbal1 to the first "slot," pverbal2 to the 2nd, ... pvisual6 to the 12th.
•     Tell SPSS that Age is a between subjects variable.
•     Tell it to plot Trials on the horizontal axis, separate lines for modality, and separate plots for age.

Now run the analysis of variance. Copy the appropriate values from the SPSS printout into the following table.