nav[aria-label="Primary Navigation"] { padding: 0; & ul { list-style: none; width: 100%; display: flex; flex-direction: row; justify-content: start; align-items: start; gap: 30px; padding: 0; & li { margin: 0; } & ul li { list-style: none; } } }

Different mixed model options

IK1

Under "Fit Model" I'm a little confused over how to run a proper Mixed Model. One approach seems to be to choose "Mixed Model" under "Personality" and then simply add fixed effects and random effects under the different tabs. However, the other option is to keep "Standard Least Squares" under "Personality" and then add variables that are then assigned as random effects using the red triangle drop-down under "Attributes". Isn't this a mixed model too? I get pretty much the same outcome when I try it with the same data these two ways, but JMP gives slightly different outputs. The latter uses REML variance component estimates. Are these the same analyses?

Accepted answers

jiancao

Thanks, @P_Bartell.

Yes, it is a mixed model, the random effects model. However, if the data is unbalanced the results are slightly different. JMP Pro's Mixed Model is preferred.

With JMP Pro you can fit other types of mixed models to repeated measures with choices of covariance structures, spatial data or random coefficient models.

All comments

P_Bartell

It's almost impossible to provide a 'one size fits all' workflow for fitting mixed models because so much depends on:

1. Which JMP product are you using? There are big differences between JMP and JMP Pro in terms of capabilities for this broad family of modeling methods. As a former JMP senior systems engineer, my stock recommendation to experimenters and modelers is if Mixed Modeling is something you really depend on, then seriously consider JMP Pro because it has a much broader set of capabilities compared to JMP.

2. So much depends on the specific experimental design structure, definition of effects, and the model you are attempting to fit. There is a multitude of pathways wrt to experimental design structure, effect estimation, etc. which all influence specific selections within the JMP/JMP Pro workflow ecosystem.

The above may not help much...but without knowing which product you are using, more details wrt to the problem at hand, the specific experimental design, etc. If you haven't already a good place to start is the JMP online documentation wrt to Mixed Models found here: Mixed Model JMP Documentation

I also recommend taking a look at this JMP On Demand webinar presented by my former colleague @jiancao with some great explanations and examples: On Demand Mixed Modeling Webinar

IK1

Thanks. Sorry, working with JMP Pro 15. I do have a specific example that I'm working through, but I won't waste people's time on the messy details. I was merely asking as a general question for the future and was hoping it was a simple "yes, they're essentially the same; use either option" or "no, they're different and here's why you would choose one over the other". However, like most things, sounds like the answer is more complicated than a yes/no! Thanks for the links. I'll watch them. I've looked at some of JMP's mixed model tutorials and training, but haven't found the issue I mentioned addressed.

jiancao

Thanks, @P_Bartell.

Yes, it is a mixed model, the random effects model. However, if the data is unbalanced the results are slightly different. JMP Pro's Mixed Model is preferred.

With JMP Pro you can fit other types of mixed models to repeated measures with choices of covariance structures, spatial data or random coefficient models.