topic Difficulty interpreting an interaction using proc mixed in SAS Procedures

Difficulty interpreting an interaction using proc mixed

stodo53 — Thu, 05 Sep 2024 08:23:50 GMT

Hi all,

I am running a mixed linear model using "PROC MIXED" procedure in SAS and it includes an interaction term. I am just having difficulty understanding the two reference groups, and what exactly they are a reference for. Would someone be able to help me interpret this? Ultimately, I would like to model the following "Compared to males (or females) on TAC, males (or females) on TAC have a higher/lower lvmi". Thanks in advance!

Current model:

	MODEL lvmi_chinali=	yearsincetx  ageattx TAC syssds_04 bmisds_r egfr_schwartz TAC*sex / SOLUTION CL;

Results:

Re: Difficulty interpreting an interaction using proc mixed

Ksharp — Thu, 05 Sep 2024 08:37:04 GMT

It is just FIXED effects. Check this:
https://support.sas.com/kb/38/384.html

Re: Difficulty interpreting an interaction using proc mixed

stodo53 — Thu, 05 Sep 2024 10:57:19 GMT

I read over this many times, and although I think I understand it, I don't really see how this could be made interpretable in laymans terms. When I add both the main effects and the interaction term, it becomes more confusing to me how to interpret this since there is only one estimate produced for the interaction.

	MODEL lvmi_chinali=	yearsincetx  ageattx TAC sex syssds_04 bmisds_r egfr_schwartz TAC*sex / SOLUTION CL;

Re: Difficulty interpreting an interaction using proc mixed

Ksharp — Fri, 06 Sep 2024 00:32:48 GMT

" since there is only one estimate produced for the interaction."
That was supposed to be , since in design matrix SEX='Male' is 1 and TAC='1' is 1 ,and only when SEX='Male' and TAC='1' it should be ,others are zero.
You'd better post it at Stat Forum:
https://communities.sas.com/t5/Statistical-Procedures/bd-p/statistical_procedures
And Calling @StatDave @SteveDenham @lvm

Re: Difficulty interpreting an interaction using proc mixed

Ksharp — Fri, 06 Sep 2024 08:34:25 GMT

Check @Rick_SAS blog:
https://blogs.sas.com/content/iml/2019/12/03/longitudinal-data-response-profile-model.html

According to Rick's blog:
sex=Male and TAC=1 is 1.0446 less than sex=Male and TAC=0