Choose Language Hide Translation Bar
Highlighted
Stephen2020
Level II

Interpreting Mixed Model

Hi All,

 

Several questions regarding linear mixed models.

1) In the "Fit Mixed" personality is there a way to get the prediction expression (Y=Bx+Zu+E), I can get it from other model personalities but can't find it in the "Fit Mixed" personality

2) I am looking at swimming performance in fish, I have two continuous fixed effects and their interaction fit against a continuous response with individual and all its interactions selected as random effects (in total 3 fixed effects:2independent and 1 interaction and 4 random effects)

          2a) when I am interpreting the random variance components is the residual in that table the variance left to be explained by the fixed effects?

          2b) If I am getting significant random variance components and significant fixed effects are my fixed effects significant BECAUSE of the random covariance            or INSPITE of the random covariance?

3) I am working with acceleration which appears to be non-linear and my model fits better when it is log transformed. Am I justified in transforming only acceleration and not my other continuous variables (which appear to be normal or nearly normal) or should I be transforming them all.

 

Here are a couple of screen grabs

Stephen2020_0-1592308261731.pngStephen2020_1-1592308312397.pngStephen2020_2-1592308383498.png

 

Stephen2020_3-1592308404348.png

 

Thanks in advance

 

1 REPLY 1
Highlighted
Byron_JMP
Staff

Re: Interpreting Mixed Model

1. From the red triangle, Save columns, save prediction formula.

expression is in a new column

2. https://www.jmp.com/support/help/en/15.1/#page/jmp/the-fit-mixed-report.shtml#

I don't see a residuals table in the attached report, however

Marginal Model Inference: Produces plots based on marginal predicted values and marginal residuals. These plots display the variation due to random effects.  https://www.jmp.com/support/help/en/15.1/#page/jmp/marginal-model-inference.shtml#ww1282958

 

Conditional Model Inference: Produces plots based on conditional predicted values and conditional residuals. These plots display the variation that remains, once random effects are accounted for.

https://www.jmp.com/support/help/en/15.1/#page/jmp/conditional-model-inference.shtml#ww1171576

 

 

3. I'm not familiar with fish acceleration. Either a log or some sort of Box-Cox transform seems like it might be reasonable.

 

JMP Systems Engineer, Pharm and BioPharm Sciences
Article Labels