# What analysis (in SPSS) should I use for a repeated measures design with dichotomous outcome variable?

Parts of my question may seem rudimentary but the more I read, the more confused I get so I am going to completely spell out my experiment.

In my study, I have created 4 conditions by crossing 2 factors- 1) comparison type 2) pricing. Each have 2 levels. My questionnaire had 12 items, 3 from each condition. MY DV is which product people choose (also 2 levels).

I understand that my DV is dichotomous and that I have a within subjects study (as all participants took part in all 4 conditions). Therefore, I cannot perform a logistic regression because my observations across conditions are not independent. I have heard people talk about Lunney's article and took a look at it myself and it seems that I can do an ANOVA. I have over 30 participants per item.

1) I have read that maybe I can use a mixed model univariate ANOVA- making comparison and pricing fixed effects and subject a random effect. Is this true? Did I understand fixed and random effects correctly in this context?

2) The other issue is that as I said, I have 3 different scores for each participant per condition, meaning 12 scores in all. How should I arrange the data? So far what I have done is made one column called "DV" and put all the responses there. I then made an ID column, putting the ID next to each response and then made 2 separate columns for the difficulty and pricing (and coded the item accordingly). Is this correct?

3) Another option I have been toying with is to use the repeated measures function in GLM. To do this, I made a composite score for each participant (an average over the three items per conditon) and put 4 separate columns for each condition. I'm not sure what I would put in the "factor" slot in the window in SPSS. Any ideas?

4) The other DV I have in this experiment is reaction time (continuous). I want to use this DV to check my manipulation that the levels of comparison type are really different. How would I go about doing this? My first thought was to use a univariate ANOVA and do something very similar to what I described above- put all the reaction times in one column and add another column with the ID as a random effect and another with the comparison level as a fixed effect.

5) In my hypothesis, I predict that all conditions with the Comparison A level will pick the cheaper product (pricing A will pick product 1 and pricing B will pick product 2). I also predict that all conditions in Comparison level B will pick product 2, regardless of pricing level. I would predict no main effects and a one way interaction. This also means that the contrast of interest would be Comparison A Pricing 1 (where I would expect participants to pick product 1) vs. the rest of the conditions (I would expect them to pick product 2). How can I do a contrast in SPSS without coding? What kind contrast options should I choose (I see there is polynomial, repeated, diff. etc.)

6) How can I check if there are differences between men and women? Just a regular chi squared test? How do I line up the data?

Thanks so much for your help!

-
Just thinking aloud- can I add in age and sex as covariance? –  Emily Jun 12 '12 at 3:23
I'm sorry for being lazy to read all your question, but... when there is a categorical DV repeatedly measured and there is also between-subject factors and covariates, the first thing what comes to mind regarding SPSS is Generalized Estimating Equations procedure. –  ttnphns Jun 12 '12 at 9:43