I ran an experiment examining how 1st and 3rd graders responded to a series of questions about events. The responses to these questions were dichotomous-Yes/No. In my study, I have created 12 conditions per age group by crossing 3 factors- 1) source who caused the event in question (3 levels) 2) suggestion about source of event (2 levels) and 3) bizarreness of event (2 levels). My questionnaire had 12 items, 1 from each condition.
So I have 12 binary (within-subject) responses to questions for each participant and 2 (between-subject) groups of participants.
I need to test whether there are main effects and interactions in the data.
If the DVs were continuous this would be a straightforward repeated measures, mixed model ANOVA.
I have been researching what analysis is best for the data and how to do it with little luck. A policy capturing approach was recommended to me previously but I have found little resources on how to do it and am uncertain if it is the best approach? I've read that there are corrections that can be made in logistic regression for overdispersion issues but was unsure if this was acceptable for repeated measures data. It seems that multilevel modeling may also be an option?
I typically use SPSS but have some experience with R. I have access to SAS and Stata and would be open to using these programs if they allowed me to do the necessary analysis. I have some experience doing logistic regression and multilevel modeling but not at this advanced of a level.
Any suggestions about what analysis might be best, what statistical software to use, and/or resources on how to do the analysis would be very welcome!