topic How do I build a Repeated Measures ANOVA model? in Discussions

How do I build a Repeated Measures ANOVA model?

HotellingWombat — Thu, 08 Jun 2023 16:29:18 GMT

Hi, I'm having trouble with statistical analyses. My dataset includes antibody data at multiple time points and I'm studying how genetic differences affect their response to vaccine. I built a repeated measures anova mixed model with unstructured covariance and it included time, age, gender, priming vaccine, 8 polymorphisms, and the crossover of these 8 polymorphisms with time, age, gender, and priming vaccine. I've run into a problem when doing posthoc multiple comparisons where "All pairwise comparisons mean-mean scatterplot cannot be shown because confidence intervals cannot be computed." Someone has suggested about the degrees of freedom being exhausted. I was wondering if anyone knows how to solve it or if I need to change my model? Thanks!

Re: How do I build a Repeated Measures ANOVA model?

MRB3855 — Tue, 14 Mar 2023 08:17:05 GMT

Welcome! It should be easy for you to verify if you have no df left for error ("exhausted" your df). Just count up your model total df's, and if that total df = N-1 (N=total sample size) then you, indeed, have no df left for error. If that is the case, your model will have to be reduced in some way; which way? That is a question that would take a more involved discussion. That said, you may have over-specified your model (e.g., too many interactions). But, as I suggest, it will take a more involved discussion to tease this apart.

Re: How do I build a Repeated Measures ANOVA model?

BriaTyler — Tue, 14 Mar 2023 08:57:12 GMT

Multiple comparison correction methods, such as the Bonferroni method, the Holm method, or the Benjamini-Hochberg method, can be used to address this problem. These methods reduce the likelihood of false positives by applying a significance level correction to each test. If you have already used one of these methods and still encounter a problem, it is possible that the problem is related to the exhaustion of degrees of freedom. In that case, you may need to modify your model and use a simpler model with fewer variables.