Multiple observations in longitudinal analysis

[GPerry1]

Hi. I'm investigating a longitudinally arrayed system with explanatory variables collected over days, and measurements of dependent variables on a monthly basis. The monthly dependent variable collection consists of collections of numerous individual observations combined to make an average, which is what I've been fitting to the longitudinally collected explanatory variables using ARIMA and AUTOREG. However, I'd prefer to use the individual observations rather than average them out for the ARIMA/AUTOREG, so that there'd be say 60 observations at each time point instead of the average. I don't have a vast programming experience, so I was wondering if anyone could suggest how to do this on the SAS platform.

[lsandell]

Hi GPerry, PROC MIXED has options to run mixed models on correlated (aka repeated measures data). Documentation here: PROC MIXED 

The syntax is similar to PROC GENMOD and you can fit continuous and categorical outcomes. The key is to have a REPEAT or RANDOM statement which tells SAS how to fit the model based on a grouping variable (like a subject ID or something similar). Hopefully this gets you going in the right direction. 

[GPerry1]

I'd considered trying it out in MIXED but I have 60 obs for these monthly observations and a current base file of 75k time entries. I don't think I can long file it, so  was hoping there were some way to bring in the individual observations like from an external file or something along that line.