topic Re: Tukey-adjusted comparisons for one location in PROC GLIMMIX in SAS Programming

Tukey-adjusted comparisons for one location in PROC GLIMMIX

wateas — Tue, 05 Nov 2024 01:47:28 GMT

I am analyzing data from a factorial experiment using PROC GLIMMIX in SAS. In the model, "Trt_Amend_App" is a categorical variable that is one of the experimental treatments and "ID_S" is one of two locations. The model results indicate that there is a significant interaction between Trt_Amend_App and ID_S such that Trt_Amend_App is significant at one location, but not the other. I would like to conduct a post-hoc test to determine which levels of Trt_Amend_App are different from each other, corrected for multiple comparisons using the Tukey adjustment, but only for the significant location "Site 2". How might I code that? I've included the code for my model below. The commented line under the first LSMEANS statement can be thought of a pseudocode indication of what I want to do (it doesn't work 🙂).

proc glimmix data=df_y3_t3 plots=studentpanel;
class ID_S Block Trt_Amend_App Trt_CC;
model Response_Var = Trt_Amend_App | Trt_CC | ID_S ID_S(Block);
lsmeans ID_S * Trt_Amend_App / slice=ID_S plot=meanplot(sliceby=Trt_Amend_App join);
/*lsmeans Trt_Amend_App / diff adjust=tukey where(ID_S='Site 2');*/
run;

I understand that the WHERE statement can be used to achieve the desired test in a separate procedure, but I'm hoping for a more streamlined solution. Thanks for reading.

Re: Tukey-adjusted comparisons for one location in PROC GLIMMIX

Ksharp — Tue, 05 Nov 2024 03:08:54 GMT

You might could try SLICE statement:
slice ID_S * Trt_Amend_App / sliceby(ID_S='Site 2') diff;

https://support.sas.com/kb/24/447.html

Re: Tukey-adjusted comparisons for one location in PROC GLIMMIX

wateas — Tue, 05 Nov 2024 18:12:48 GMT

Thanks for the suggestion, but I don't see any additional output when I add the statement.

Re: Tukey-adjusted comparisons for one location in PROC GLIMMIX

Ksharp — Wed, 06 Nov 2024 00:26:07 GMT

Maybe I don't understand what you are looking for very well.
Post your question at Statistical Forum:
https://communities.sas.com/t5/Statistical-Procedures/bd-p/statistical_procedures

@StatDave @SteveDenham might give you a hand.