How do I make my Mixed model work with AR(1) covariance structure?

AvgMethodMoose7 — Thu, 01 Aug 2024 18:56:01 GMT

I have a large(2010X 4) data set and I need to calculate the effects of cycles and stresses on my wave speeds along with the interaction between them. My wavespeeds are collected for 4 stress levels so stresses here are repeated measurements. Here is a glimpse of what my stacked table looks like : N=subject on first row, CYC=cycle number on second row, STR: stress levels on third row and wave speeds collected in fourth row

I used the Mixed model on JMP pro where I have my cycle and stresses full factorial on fixed effects, N(subject number) on random effects, and wave speeds in Y.

My question here is in repeated structure, I am unsure whether I should use a residual structure or AR(1) for repeated covariance structure. when I use residual structure for my co variance these are my results :

In the fixed effects, the cycles, and stresses individually shows significance however cyc*stress interaction does not show significance which is not what is interpreted . and the cycles Nparm shows 43 which is taking account of only one fascic How do I approach this problem ?

Re: How do I make my Mixed model work with AR(1) covariance structure?

MRB3855 — Mon, 05 Aug 2024 10:23:15 GMT

Good question @AvgMethodMoose7 : There are many ways of thinking about this, but here is a good reference (earlier version of JMP, but the same thinking applies).

https://community.jmp.com/t5/Mastering-JMP/Fitting-Repeated-Measures-Data-using-JMP-Pro/ta-p/312501

Pretty good article here as well.

https://support.sas.com/resources/papers/proceedings/proceedings/sugi30/198-30.pdf

topic How do I make my Mixed model work with AR(1) covariance structure? in Discussions

How do I make my Mixed model work with AR(1) covariance structure?

Re: How do I make my Mixed model work with AR(1) covariance structure?