|Date: ||Fri, 24 Jan 2003 13:56:26 -0500|
|Reply-To: ||"Park, Ju-Hyeong" <gzp8@CDC.GOV>|
|Sender: ||"SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>|
|From: ||"Park, Ju-Hyeong" <gzp8@CDC.GOV>|
|Subject: ||Non-positive definite estimated G|
I am analyzing variance component with a random model using Proc mixed. But
I have been facing "nonpositive definite estimated G" matrix issue when I
ran the model with one random effect and a fixed intercept. And I got zero
variance for the random effect. Is there any way to overcome this issue?
Any comments are very appreciated.