05-22-2014 09:19 AM
here are the examples:
RANDOM <my_variable1> / RESIDUAL SUBJECT = <my_variable2>;
RANDOM _RESIDUAL_ / SUBJECT = <my_variable2>;
05-23-2014 10:10 AM
The former syntax specifies an R side covariance structure for <my_variable1>, while the latter syntax models an overdispersion. I believe that since the default type is TYPE=VC, the two statements as they currently stand should yield the same results. If not, please post, as this is a potentially interesting event.
05-26-2014 01:35 PM
Steve, at first glance the outputs are the same. I also met another explanation by @lvm:
If you do not have missing values for trial (i.e., same trials for each subject), you can use:
RANDOM _RESIDUAL_ / SUBJECT=SUB TYPE=CS;
Probably this is the same you mention.