PROC MIXED FOR POOLED REGRESSION?

[Dipanjan]

Hi

I have a data of structure as 6 subjects each having 36 time point observations. Each of 6x36 rows have first column for a continuous response variable and the rest 3 columns for three continuous predictor variables. I need to build a model for each subject having different intercept but same slopes. Obviously there are issues of perfect link function, estimation method etc but those are of secondary interest.

One way I found is to add 6 binary predictors for 6 subjects and run PROC GLM to have different intercepts with same slopes for each subject. But probably this is considering the effects of states to be fixed not random. I read about usage of PROC MIXED in this article http://cba.ua.edu/~mhardin/ST610/ST610Web/LongitudinalAnalysisPapers/LONGITUDINAL.pdf and it seems that here we can use fixed as well as extend to random effects too. But I am not sure whether it can consider subjects effects to be fixed or random. If it considers so then can we get different intercepts for different subjects?

If PROC MIXED is not the solution then can anyone suggest another procedure that fulfills my requirement ?

Thanx

Dipanjan

[Jack2012]

I am interested in such kind of questions, but because I know little about MIXed model, so I am waiting someone can help to solve this good question.