How to analyse multiple between subjects factors in a mixed ANOVA in SPSS?

Question

I am performing a Mixed between subjects Anova where I have multiple time points that subjects have completed a measure at.

• The within subjects factor is time.
• The between subjects factors include gender, ethnicity, refugee status and English language.

My two questions are

1. Combined or separate ANOVAs: Is it possible in SPSS to do this as one analysis or do I need to do multiple Anovas one for each of my between subjects factors? If it is possible to do it in one model can you explain how specifically in SPSS?
2. Fixed versus random factors: My advisor told me to be sure to know whether I am using the models as fixed effects or random effects but she herself doesnt really understand this. I understand the difference ie fixed effects are like gender versus random effects that are random sample of the total # of levels in the population however I am not sure how to deal with this in SPSS or even tell what the default is. I performed my analyses using Analyze, General Linear Model, Repeated Measures.

Answer 1

Combined or separate ANOVAs

• You probably want to include all the predictors to develop an overall model. By default SPSS will introduce all possible interactions between your predictors. I doubt you would be interested in testing whether there is a gender by ethnicity by refugee status interaction. As a starting point I'd click on the "Model" button and select only the effects of interest (e.g., main effects, perhaps interactions between between subjects factors and time; perhaps others). * Also by default SPSS REPEATED MEASURES will treat time as an unordered factor, whereas often it is better treating time as a numeric variable where you explore linear, quadratic or other functional effects of time. Polynomial contrasts are one simple option to explore in SPSS REPEATED MEASURES.
• If you want to get more sophisticated then you might want to explore multilevel modelling options which provide more flexibility.

Fixed versus random

• All your between subjects factors are presumably fixed factors in that you have measured all levels of the factor of interest, or at least you have not sampled your factors from a domain of possible factors.