Help With PROC GLIMMIX Error Messages

Nebulus — Thu, 16 Jan 2020 20:52:34 GMT

So I'm running an experiment where I am taking counts of insects. I have 3 treatments with 10 replicates per treatment arranged in a RCBD (There are 10 blocks containing each treatment). I take repeated counts for 6 consecutive weeks. I replicate the entire experiment twice. I'm treating weeks as a repeated measure and block and experimental replicate as random effects.

Here is my code. I'm getting the dreaded "Obtaining minimum variance quadratic unbiased estimates as starting values for the
covariance parameters failed." error and not sure where the error lies in my coding. For sake of simplicity I'm just showing the model looking at the variable "trips_r" for now but I will be looking at 8 different insect count variables and the damage variables (poprate and roserate). (TRT = treatment, EXP= experiment replicate, WEEK= week, CAGE= experimental unit, BLOCK = block).

proc sort;
by TRT EXP WEEK;
RUN;

PROC PRINT DATA=greenhouse;
run;
quit;

PROC MEANS stderr noprint; by TRT EXP WEEK;
var adultm_p eggm_p totm_p trips_p adultm_r eggm_r totm_r thrips_r peprate roserate;
output out=avg
mean = Xadultm_p Xeggm_p Xtotm_p Xtrips_p Xadultm_r Xeggm_r Xtotm_r Xthrips_r Xpeprate Xroserate
stderr = Sadultm_p Seggm_p Stotm_p Strips_p Sadultm_r Seggm_r Stotm_r Sthrips_r Speprate Sroserate;
run;

proc print data=avg;
run;

PROC GLIMMIX DATA=greenhouse;
CLASS TRT WEEK EXP BLOCK CAGE;
MODEL thrips_r = TRT|BLOCK|EXP / dist=poi link=log s;
random intercept / subject=BLOCK;
random WEEK / type=ar(1) subject=CAGE(TRT*BLOCK) residual;
LSMEANS TRT/ PDIFF adjust=TUKEY;
RUN;

Re: Help With PROC GLIMMIX Error Messages

Haris — Sat, 18 Jan 2020 14:56:07 GMT

This is not a code error. Most likely, you don't have enough data to estimate desired parameters given the complexity of your model. Alternatively, there is not enough variability in your data for the algorithms to do their job.

Solutions would be:

- Simplify or change your model.

- Get more data.

- Introduce more variability.

topic Help With PROC GLIMMIX Error Messages in Statistical Procedures

Help With PROC GLIMMIX Error Messages

Re: Help With PROC GLIMMIX Error Messages