Linear Mixed Models in SPSS: why the extra repeated measures option??

#1
Hi all,

I've got a basic question about the interface of SPSS linear mixed models.
The initial window provides the option to designate one or more variables as "repeated measures" and to choose a separate covariance structure for these.

I noticed that not using this option doesn't do an awful lot to the parameter estimates, but there is a slight difference (seems to slightly increase the p values, although hardly noticeable).

Just wondering: why is this option given?
Isn't it enough to put the repeated measure variables in the random effects field?

Appreciate any thoughts on this!
Bertram
 

Dason

Ambassador to the humans
#2
I've never used SPSS but there is a difference between the two options. For instance if you just take care of repeated measures by incorporating a random effect this forces a compound symmetry structure on the covariance matrix between observations that share the same random effect. However, sometimes it might be more appropriate to use a difference covariance matrix - an AR(1) correlation structure for instance might be more appropriate for measures that are repeated over time.