topic Trouble with multiple interaction term in Proc Mixed Repeated Measures in Statistical Procedures

Trouble with multiple interaction term in Proc Mixed Repeated Measures

deleted_user — Thu, 29 Apr 2010 19:16:34 GMT

I am trying to run a repeated measures model in proc mixed with relative seed production as my dependent variable and year (the repeated measure), genotype, and size as independent variables for each plant in my study. The model runs well if I include a single interaction term such as year*size, but when I try to include year*genotype*size I get the "stopped because of infinite likelihood error." My code is below. I've tried using type =un, type= AR(1) and type = cs, but nothing works. Any suggestions?

proc mixed;
class genotype plant year;
model rel_seed_prod = year genotype size year*genotype year*size genotype*size year*genotype*size;
repeated year / subject = plant type = un;
run;

thanks for your help!
Carolyn

Re: Trouble with multiple interaction term in Proc Mixed Repeated Measures

Doc_Duke — Fri, 30 Apr 2010 00:38:21 GMT

Have you looked at
http://support.sas.com/documentation/cdl/en/statug/63033/HTML/default/statug_mixed_sect028.htm
? My guess is that you've got too many parameters for the amount of data that you have and the Newton-Raphson search algorithm is failing to converge.

Re: Trouble with multiple interaction term in Proc Mixed Repeated Measures

Dale — Fri, 30 Apr 2010 18:50:12 GMT

Carolyn,

You might try initializing the covariance structure parameters according to the values which were estimated when you have only the year*size interaction in the model. Having reasonable initial estimates of the parameters of the covariance structure can help one avoid some estimation problems.