topic Re: Random effects distribution in proc glimmix in SAS Procedures https://communities.sas.com/t5/SAS-Procedures/Random-effects-distribution-in-proc-glimmix/m-p/216619#M53332 <HTML><HEAD></HEAD><BODY><P>The random effects can only have a normal distribution in glimmix (of course, the response variable can have any of the conditional distributions in the exponential family). Random effects must also be normal in nlmixed (no other ones allowed).&nbsp; With nlmixed, there are some tricks to get other random-effect distributions (I have tried some), but these are very tricky to use, and often do not converge. Fortunately, most model fitting results are not too dependent on the normal assumption for the random effects. </P></BODY></HTML> Mon, 06 Apr 2015 19:33:06 GMT lvm 2015-04-06T19:33:06Z Random effects distribution in proc glimmix https://communities.sas.com/t5/SAS-Procedures/Random-effects-distribution-in-proc-glimmix/m-p/216618#M53331 <HTML><HEAD></HEAD><BODY><P>Hi SAS Community,</P><P>I have a quick question about proc glimmix. When using proc nlmixed, the analyst specifies the distribution of the random effects. Is there any such option with proc glimmix? I could not see. If there is no such option in proc glimmix, can anyone please suggest a way to specify a random effects distribution different from the normal distribution? Thanks.</P></BODY></HTML> Mon, 06 Apr 2015 16:11:20 GMT https://communities.sas.com/t5/SAS-Procedures/Random-effects-distribution-in-proc-glimmix/m-p/216618#M53331 Oms 2015-04-06T16:11:20Z Re: Random effects distribution in proc glimmix https://communities.sas.com/t5/SAS-Procedures/Random-effects-distribution-in-proc-glimmix/m-p/216619#M53332 <HTML><HEAD></HEAD><BODY><P>The random effects can only have a normal distribution in glimmix (of course, the response variable can have any of the conditional distributions in the exponential family). Random effects must also be normal in nlmixed (no other ones allowed).&nbsp; With nlmixed, there are some tricks to get other random-effect distributions (I have tried some), but these are very tricky to use, and often do not converge. Fortunately, most model fitting results are not too dependent on the normal assumption for the random effects. </P></BODY></HTML> Mon, 06 Apr 2015 19:33:06 GMT https://communities.sas.com/t5/SAS-Procedures/Random-effects-distribution-in-proc-glimmix/m-p/216619#M53332 lvm 2015-04-06T19:33:06Z Re: Random effects distribution in proc glimmix https://communities.sas.com/t5/SAS-Procedures/Random-effects-distribution-in-proc-glimmix/m-p/216620#M53333 <HTML><HEAD></HEAD><BODY><P>Thanks, Ivm.</P></BODY></HTML> Tue, 07 Apr 2015 04:05:53 GMT https://communities.sas.com/t5/SAS-Procedures/Random-effects-distribution-in-proc-glimmix/m-p/216620#M53333 Oms 2015-04-07T04:05:53Z