Your code generates a warning (in the log) and odd results because you have specified Group as a fixed effects factor in the model, and in the data step you have specified Group effects of exactly zero. If effects of Group were instead nonzero, then the code would run. But this code is for a two-way factorial (without interaction) in a completely randomized design, and you intend your design to be a crossover.   Replication in a crossover design can take different forms. For example, subjects could be blocked in sets of 3 ("squares"), or subjects could be independent (i.e., not clustered or blocked). Your power analysis needs to mimic your intended design. How do you envision doing replication in your study?   Once you know how you plan to replicate (and need only to figure out how many replications), then it may be helpful to write the code (which I will call the "data analysis code") for the experimental design that you anticipate. You can then compare the model in your data analysis code to the model in your power analysis code: if the models do not match, then you need a different power analysis specification.   You most likely will not be able to do power analysis for a crossover design with GLMPOWER because a crossover design is fundamentally a mixed model, and GLMPOWER does not accommodate mixed models. In the SAS Community thread that I linked to in my previous message, pay particular attention to the response by lvm who references a text and a paper by Walt Stroup; I referenced a second paper in my previous message in this thread. You also could study Ch 12 in https://www.sas.com/store/books/categories/usage-and-reference/sas-for-mixed-models-second-edition/prodBK_59882_en.html. Stroup's approach takes advantage of the ability to hold variance parameters constant (PARMS ... / HOLD) and does not require simulation of data. It is the approach that I use to estimate sample size for a mixed model.   I hope this helps.    Edit: Note also that your CONTRAST statements are incorrect. By default, the order of levels for Treatment is alphabetical: Drug1, Drug2, Placebo. Your coefficients need to match the order of levels.   ... View more

Growth chamber studies are challenging to design, largely because no one has a whole bunch of growth chambers to work with. Or sometimes even more than one.   I'd probably use the model I suggested and be sure to make the experimental design clear in Methods. The risk you run with this design is that if there was an effect associated with the 35/25 treatment, then it could be due to the treatment per se or it could be due to that particular growth chamber. Or both. You have no way to know for sure.   ... View more

You have 4 levels of temperature and 3 growth chambers, so you clearly did not assign the same temperature level twice to the same growth chamber; if you did to some extent, then the models I propose below are not consistent with your design.   If we assume that there is no systematic chamber effect on the mean response and that there is no systematic temporal effect on the mean response, then we can think of your design as having 8 growth chambers, randomly assigned to 4 temperatures. (This obviously is not what you did, but might be a good-enough model if the assumptions are valid.) The 3 blocks within each chamber are not true replications, rather they are subsamples because they are not independent realizations of a particular temperature level. You can fit a model with the full set of 144 observations:   proc glimmix data=wue plots=(studentpanel boxplot(fixed student)); class run block temp var; model inswue = temp var temp*var / ddfm=kr2 ; random run(temp) block*run(temp) var*run(temp); lsmeans temp var / diff adjust=tukey lines cl; lsmeans temp*var / plot=meanplot(sliceby=var join cl) slice=(temp var); run; or you can compute the means over the 3 blocks within each chamber for each variety and fit a model with 48 observations:   proc sort data=wue out=wue_srt; by run temp var; proc means data=wue_srt noprint; by run temp var; var inswue; output out=wue_means mean= ; run; proc glimmix data=wue_means plots=(studentpanel boxplot(fixed student)); class run temp var; model inswue = temp var temp*var / ddfm=kr2 ; random run(temp); lsmeans temp var / diff adjust=tukey lines cl; lsmeans temp*var / plot=meanplot(sliceby=var join cl) slice=(temp var); run; This second model is a two-way factorial in a split-plot design; the first model is a form of split-plot with subsamples. Because you have the same number of blocks in each chamber, the test results are the same for both models.   I think these models are, more or less, valid and certainly more correct than your original model, if I correctly understand your design. There is no evidence of interaction, so (unfortunately) your original question is now moot.   I hope this helps.       ... View more

As noted in Tips and Strategies for Mixed Modeling with SAS/STAT® Procedures by Kiernan et al., this warning could be due to an overspecified model. Try omitting   random intercept / subject=person_id; This paper by Littell et al. describes the difference between an AR(1)+RE structure and an AR(1) structure.   The Kiernan et al. paper lists other potential solutions to this problem.   I'd also add the <repeated-effect> to the REPEATED statement; it may not be necessary for your data but it's a programming safeguard. See the documentation.   I hope this helps.   ... View more

Your RANDOM statement looks odd to me; it does not correspond to any experimental design that I would consider to be likely. I think it probably is not correct but I would need details about your experimental design to say for sure.   ... View more

If the overall test of interaction is significant, but the simple effects (as obtained by, say, the SLICE option on the LSMEANS statement) are not, then I typically take this as an indication of appreciable uncertainty about the existence of interaction in this particular set of data--in other words, I may not be convinced that interaction exists, or I might think interaction probably exists but I don't have much as much evidence as I would like for that conclusion.   By removing the interaction term from the model, I am explicitly assuming that there is no interaction. Maybe I'm comfortable with that assumption, or maybe I'm not. It depends on the scenario.       ... View more

Does that mean that you have only two values for each HHID, namely stuntingall3 and peduc? If so and if you are using stuntingall3 as the response and peduc as a predictor, then you do not have a data structure with repeated measures and the REPEATED statement is not necessary and would generate errors.   Looking ahead, if peduc has only two levels, then the ESTIMATE statement is not correct because it specifies 3 parameters.   ... View more

A crossover is a mixed model of some complexity, and I don't think any of the power analysis PROCs address it directly.   I'll point you to this SAS Community thread. Basically, there are two approaches: simulation and a nice trick discussed by Walt Stroup in his texts and papers. Here's a more recent paper by Walt.   This thread is a few years old, and Rick Wicklin's book on simulation is now published.    With either approach, once you've done the power analysis, you now have code already written for the analysis of actual data. Sweet.   I hope this helps.   ... View more

Tech Support has been helpful with all of my questions in the past--hopefully you'll have the same experience. Let us know!   ... View more

https://support.sas.com/en/technical-support/contact-sas.html   shows the different options.   Good luck!   ... View more

I regret that I am not able to figure out your design structure from the additional information that you've provided.   I doubt that two blocks are defined by the two conditions. Blocks are typically considered to be random effects factors which determine the variance of the response and do not affect the mean of the response. I believe you are thinking of condition as a fixed effects factor; if condition is a fixed effects factor, then it cannot also be a random effects factor. So you'll want to think more about your design structure.   Blocks are clusters of experimental units: units within a particular block are assumed to be "similar", and units between blocks are assumed to be "different". You may have clustering in your experimental design, or you may not.   Design issues can be complicated, so if there is someone at your institution who could help, I suggest talking with them. There is, of course, a lot of books, papers, etc. on the topic of design.   Or you could try again here, with a lot more detail about how you ran the experiment.    ... View more

Proc MODEL is capable of multiple regression, but it has no CLASS statement. Theoretically, you could code indicator variables for FIRM and YEAR and then use those indicator variables in the MODEL statement. But 5000 FIRMs? I have no idea whether that is possible. I would ask SAS Technical Support.   ... View more

To apply Newey-West, the clearest path is to switch to PROC MODEL in SAS/ETS. See Usage Note 40098: Newey-West correction of standard errors for heteroscedasticity and autocorrelation   I doubt that Newey-West is possible using PROC GLM.   ... View more

I would say that your treatment structure is a two-way factorial: extraction method (4 levels) and condition (2 levels). But you have not told us anything about your design structure so we have no way of knowing whether the design is a CRD, or a RCBD, or a split-plot, etc.   If you do an internet search on "experiment treatment structure design structure", you'll find several helpful links.   I hope this helps.   ... View more