Random and Repeated Statement in MIXED

[Ryan]

Hello:

I was asked recently in a private email about the validity of adding a REPEATED statement to the MIXED code I posted a while back to fit a multivariate model. My original post can be found here:

http://old.nabble.com/Multivariate-Analysis-in-MIXED-to27027781.html#a27027781

The person asked if a REPEATED statement could be added to the model. If we stick with the example I presented previously (2 observations per person),  the answer is definitely no, regardless of what "Indic" represents.

Now, if "Indic" represents a variable such as "time" and there are more than 2 observations per person, then a REPEATED statement in addition to the RANDOM statement may be appropriate. The REPEATED statement would like this:

REPEATED = Indic |subject(ID) covtype({UN or AR or...})

The random statement would account for person-specific intercepts, and conditional on the person-specific intercepts, the REPEATED statement would account for residuals being correlated based on proximity in time. If the var-cov matrix for the REPEATED statement is compound symmetric, then the REPEATED statement should be excluded.

HTH,

Ryan