Sample size estimation for mixed model designs

bio_grad

Hello,

 

I am somewhat new to the concept of mixed model analyses for repeated measures but I believe that they will be of great use to my projects. One thing that I am having trouble grapping with is how to go about estimating what type of sample size I should be aiming for.

 

ANOVA and t-test sample size calculations are easy to follow but I don't even know where to start for mixed model repeated measures.

 

I have some pilot data as well as past literature but these were much simpler studies. I have an idea of variation and effect sizes but no idea how to apply them in the context of mixed models.

 

I feel as though simulation might a potential option to look at. Wondering if JMP Pro 13 has this capability.

 

Thanks!

Kevin_Anderson

Hi, bio_grad.

 

JMP has any capability that you have the motive and opportunity to script into it.  Your questions can certainly be addressed with some (probably rather grueling) JSL and some (definitely grueling) detailed modeling knowledge. However, based on what you've said, you may want to reconsider what you're getting yourself into.

 

But a quick online search found some interesting information that would seem germane to me, a casual observer.

 

https://com-hop-sample-size-shop.sites.medinfo.ufl.edu/files/2012/08/Easy-Power-and-Sample-Size-for-Most-of-the-Mixed-Models-You-Will-Ever-See.pdf

 

And therein referencing GLIMMPSE at the SampleSizeShop.org, which looks very interesting upon first glance.

 

I assume you have done a search before questioning the Discussion Forum.  Do these not suit your needs? 

 

I also recommend that you seek the professional help of an experienced statistician with whom you can consult face-to-face.