Both approaches can deal with the structure of your data. The random effects model in MIXED is a subject-specific model best for individual predictions. The GEE model is a marginal or population-averaged model that is best for making population inferences.  But as noted by Allison in his book, "Fixed Effects Regression Methods for Longitudinal Data Using SAS"  (Allison, P., SAS Institute, 2005), these are effectively equivalent in the case of the linear model like yours, though you might want to use the exchangeable structure (TYPE=EXCH) in the GEE model. Another possible approach is the fixed effects model that Allison's book also discusses and which is implemented by the ABSORB statement in PROC GLM.    Note that the recommended procedure for fitting the GEE model is now PROC GEE though GENMOD can certainly be used. Also, the GEE model does not use a likelihood-based approach, so model comparisons using the likelihood or measures like AIC are not possible. ... View more

See this general note on scoring. Try method 3. ... View more

First, your model specification is not producing the separate main effects and 2-way interactions. It only contains the 3-way interaction. However, all the degrees of freedom of the main effects and 2-way interactions are included in the 3-way interaction using that specification. So, it is effectively equivalent to writing      model PercStemEndRot=Harvest|Variety|Wks which is the shorthand way to specify all of the main effects and interactions.   But the bottom line is that you'll need to simplify the model because the model complexity using all the effects (either explicitly or implicitly specified as noted above) is making the data too sparse. I suggest that you start with only the main effects model to see if it is successful:     model PercStemEndRot=Harvest Variety Wks and then add interactions one at a time as long as the fit succeeds - initially without FIRTH and then adding it if needed. ... View more

CATMOD does not fit GEE models. The recommended procedure is PROC GEE though it can also be done using PROC GENMOD which is what the Margins macro uses.   See this note which uses a GEE modeling approach to interrupted time series and allows for nonnormal response distributions like the negative binomial distribution. The negative binomial GEE model can handle both the repeated measures nature of your data and overdispersed count data such as caused by excessive zero counts. The note shows examples of studies with one group or more and with normal and binary response data. While an example with count data is not shown, it would be done essentially like the binary response examples but with selection of a different distribution and link function. Links at the beginning of this note point to related notes that might be of interest, particularly the one about spline effects which might be useful to you if your trajectories over time are not well approximated by simple linear or polynomial effects in the model. The last example in the note would be close to the study design that you mention. ... View more

You can still use GLIMMIX if you don't have random effects - just don't include a RANDOM statement. Again, see the usage note I referred to. For your continuous proportions, you can fit the fractional logistic model using either GLIMMIX or LOGISTIC as shown there. In the example there, the RANDOM statement in GLIMMIX is used just to estimate a reasonable scale parameter (it does not add a random effect) in the same way as the SCALE=PEARSON option is used in PROC LOGISTIC in that same note. Both use the binomial distribution. But as I've said, if the data in each separate population exhibits a distribution that is reasonably normal, then an analysis using that distribution might be reasonable. You might want to try both and see what trade-offs there. ... View more