Exponentiate the equations in the estimate statements:

estimate '2010' exp(b0 + b_ACFL*0.33 + b_INBU*0.33 + b_forest*0.33 + b_field*0.33 + b_year*2010);

Robin High UNMC

-----Original Message----- From: SAS(r) Discussion [mailto:SAS-L@LISTSERV.UGA.EDU] On Behalf Of Andrew Cox Sent: Thursday, March 22, 2012 3:16 PM To: SAS-L@LISTSERV.UGA.EDU Subject: Back-transforming a truncated poisson parameters in NLMixed

Hi, I stole some code from the SAS website to model some data that fit a truncated poisson distribution (no zero values). Here is the code I used, which (I hope) models intensity as a function of several variables and includes a random effect for the location each sample was taken from.

proc nlmixed data=nlmixeddata; parms b0=0 b_ACFL=0 b_INBU=0 b_Forest=0 b_field=0 b_year=0 s2u=0.5; lambda=exp(b0 + b_acfl*acfl + b_inbu*inbu + b_forest*forest + b_field*field + b_year*year + u); ll=-lambda+intensity*log(lambda)-lgamma(intensity+1)-log(1-exp(-lambda)); model intensity ~ general(ll); random u ~ normal(0,exp(2*s2u)) subject=location; estimate '1991' b0 + b_ACFL*0.33 + b_INBU*0.33 + b_forest*0.33 + b_field*0.33 + b_year*1991; estimate '2010' b0 + b_ACFL*0.33 + b_INBU*0.33 + b_forest*0.33 + b_field*0.33 + b_year*2010; run;

What I don't understand is how to back-transform the output from the estimate statements so it is in the same scale that the original intensity variable is.

Any help is much appreciated.

Andrew Cox University of Missouri