topic Re: Help with plotting data from a mixed model design in Statistical Procedures

Re: Help with plotting data from a mixed model design

UnkaGee — Thu, 09 Apr 2020 13:01:42 GMT

Great point Paige and thanks for your quick reply.

I think a decent way to visualize this would be by group. So ideally, I would have 2 graphs (1 for NSA inspectors and one for TSA inspectors). For the axes, response time would be the y-axis (DV) and x-axis would be (IV) # items 2, 4, 8. The line connecting the points would be solid for the "yes" condition and dotted for the "no" condition.

Thanks for the referral to STORE and the standard GLM plots, I will give them a look.

Re: Help with plotting data from a mixed model design

Rick_SAS — Thu, 09 Apr 2020 19:40:00 GMT

I recommend using the EFFECTPLOT statement in PROC PLM to visualize regression models in SAS.

I also suggest PROC MIXED for repeated measures. Some EFFECTPLOT examples are at this blog post,

although the examples only have one continuous explanatory variable. You have to combine the ideas in the two posts.

Re: Help with plotting data from a mixed model design

SteveDenham — Fri, 10 Apr 2020 14:50:50 GMT

@Rick_SAS mentions using PROC MIXED for repeated measures. There is a good reason for that. PROC GLM does NOT use (or even compute) the correct error terms for any whole plot factors. See SAS for Mixed Models (I think it is on 3rd edition now).

SteveDenham