In the MIXED procedure, the TYPE=AR(1) covariance structure specified in the REPEATED statement is designed for repeated measures data for which the repeated measures are taken on the same set of equally-spaced time points for each subject. It is recommended that you always specify the repeated effect in the REPEATED statement in PROC MIXED.

If your design has unequally-spaced time points, or if your subjects are observed over differing sets of time points, then consider using the TYPE=SP(POW)(time) structure instead. The SP(POW) structure models the strength of the correlation between two observations as a function of the distance between the two observations. Points closer in time will be more strongly correlated than points farther apart in time. The SP(POW) structure produces the same results as AR(1) for equally-spaced time points for your subjects.

PROC MIXED can model a variety of spatial covariance structures. For details, see the description of the TYPE= option in the "Syntax: REPEATED Statement" section of the MIXED documentation.

Operating System and Release Information