Prev: %Squeeze Change 9.1.3 to 9.2
Next: How to call ORACLE function via SAS (via metadata server) not SQL pass-through
From: Bijay Tamang on 11 Nov 2009 18:02
I am using the following code to analyze my count data. "Nxtot" is the
number of roots. When I use the following code, some of the variables and
their interactions are significant but residuals are not normal. If I use
"Dist=lognormal link=identity" in the following code, none of the variables
and their interactions are significant but the residuals are normal. AIC
values are below 300 when dist=lognormal is used but above 1000 when
dist=poisson is used.

Do residuals need to be normal in GLIMMIX?

PROC GLIMMIX data=nx4 method=quad plots=all;
class wba location;
MODEL nxtot = wba depth1 bd/DIST=poisson LINK=log SOLUTION ddfm=bw;
random _residual_;
RUN;


Thanks
 | 
Pages: 1
Prev: %Squeeze Change 9.1.3 to 9.2
Next: How to call ORACLE function via SAS (via metadata server) not SQL pass-through