Constraining the correlation of two random effects in proc MIXED

[Eran]

Hi all,

Does anybody know how I can constrain the correlation between two random effect to be 1?

I know about the PARMS statement (with the hold option), but with this statement I can constrain the Co-variance, but is there a direct way to constrain the correlation itself?

Thanks a lot

Eran. 

[SteveDenham]

I don't know whether this will work, but it is possible to specify the covariance structure as being based on the correlations, using type=unr.  From there, it seems possible that you could constrain a correlation to be a fixed value.

Steve Denham