You are running with the default RSPL method, which often has trouble with binary data. Here is my recommendation of how to change your code:   proc glimmix data=home.FS_Binary method=laplace; NLOPTIONS maxiter=1000; class RespID pant_num group_num time DemAge00; model HHFS_bi(ref="0")=group_num time DemAge00 group_num*time / dist=binary link=logit solution /* ddfm=kr*/ oddsratio; /* since we added PartID and site in the random statement, they are considered as random effect and do not need to be included in the model */ random pant_num/subject=respID; lsmeans group_num*time /oddsratio ilink e;/*see the position of the levels below*/ estimate 'intervention T2-T1 - control T2-T1' group_num*time 1 -1 -1 1/ exp cl; run; This shifts to a maximum likelihood integration method (Laplace) which doesn't allow Kenward-Rogers degrees of freedom. I also suggest rewording the RANDOM statement to get processing by subject. However, based on the sample data, it appears that each pant_num has only a single subject. If that is the case for the full dataset, you'll need to replace pant_num with intercept.   SteveDenham   ... View more

I think @Ksharp had it right the first time - PROC GLM or NPAR1WAY. No need for a mixed model here. There are 31 experimental units (the measurements) where each received one of three treatments (J, K or A). This code tests for differences between recorders using PROC GLM:   proc glm data=yourdata; class recorder; model value=recorder; means recorder/lsd lines hovtest; /* you may want to substitute tukey for lsd */ lsmeans recorder/pdiff; run; quit; And this for the nonparametric analysis:   proc npar1way data=yourdata; class recorder; var value; run; This gives several different approaches, but the easiest to consider is the Kruskal-Wallis test. The global test is not significant, so no post hoc pairwise analysis is done.   Now for the caveats: Of the 31 observations, 26 are for recorder J, 4 are for recorder K and 1 is for recorder A. This imbalance makes comparisons difficult, especially with recorder A. Variability is essentially set by recorder J.  None of the tests for difference are significant but this does NOT mean the recorders are equivalent. It may just mean that there is insufficient data to detect a real difference.   SteveDenham         ... View more

Rather than a ratio of each component to the total you might consider a ratio of consumption of the feed with the larger feed level to the smaller feed level for each pen. You get a single value that is indexed by the variable 'Treatment'. Or consider the difference in consumption calculated in much the same way. That might even be better as it is much more likely to collectively have residuals that are distributed normally. The lsmeans by treatment are then a measure of how much more was consumed of the higher valued feed - which is a clear measure of preference.   SteveDenham  ... View more