I need to analyze some data derived from an experiment in which I'm comparing two groups (independent factor) at three different time points (first repeated level) but, for each of these time points I'm measure time bins (5 bins per time point). So, in other words I have two levels of repeated measures. How can I analyze this on SPSS via Linear Mixed Model? What I mean is, how do I tell to SPSS that the bin factor is "within" the time point factor? Maybe I should use GLMM? I hope I was clear enough.
I have only used SPSS to analyse multilevel data with crossed random effects, so take this with a grain of salt, but to the best of my understanding the syntax would go like this (SPSS is VERY unintuitive re: mixed models):
MIXED dependent BY group /FIXED=group | SSTYPE(3) /METHOD=REML /RANDOM=INTERCEPT | SUBJECT(subj1) COVTYPE(VC) /RANDOM=INTERCEPT | SUBJECT(time) COVTYPE(VC) /RANDOM=INTERCEPT | SUBJECT(time*bin) COVTYPE(VC).
(Subj1 represents your participant/measurement unit id). The last row would give you what you need. It looks like an interaction between random effects, but the SPSS support page suggests that it in fact represents a nested random effect:
I also cross-checked using mock data and running the analysis in SPSS using the above syntax and in R (where specifying nested random effects is very easy) using
model<-lmer(dependent ~ (1|subj1)+(1|time/bin)+group, data=df)
and SPSS, with the above syntax, and R, with the above code, produced the same results (up until 5th or 6th decimal, there are some differences in mixed model procedure between SPSS and R which account for that). Still can't swear this is the correct syntax but this is the closest I could come up with.
Edited to add that the SPSS help page says the "interaction = nested random effect" refers to GLM and unianova but it does seem to me that this is also how it works in the Mixed syntax.