Multigroup Analysis

Question

Hello everyone,

I am new to JMP and I wonder if there is a capability to do multigroup analysis after specifying and identifying the model. Our model includes many latent variables.
An identical survey has been used to collect data from three different panels for three different study conditions. We would like to understand whether there are significant differences in effect sizes, path coefficients, and other test statistics between three groups for the same model.

Thank you.

SDF1 · Accepted Answer

Hi @McMaster ,

Yes, JMP can do this. Not knowing what the structure of your data table is, it's hard to provide more specifics, but if you already know what the model is and all the hidden variables, then it should be pretty straight forward.

If your prediction formula for the response is easy, you can make extra prediction columns and swap out the Group factor in the column formula. If it's rather complex with many terms, you might want to re-run the model in the Fit Model platform and just swap out Group A, B and C each time, but keeping all the other terms in the model the same. You'll then get three different response model fits that you can do all your comparisons with.

Or, and this would depend on the data structure of your table, you can use the Group (levels A, B, C for example) in the "By" role of your Fit Model platform. It will then run the model of your choice for each of the levels of Group. Take a look at the Diabetes.jmp file and use the continuous Y for the response and then the Binary or Ordinal Y as the By variable and select Age through Glucose as the model effects. When you run this model, it will create a new model fit for each level of the Binary or Ordinal Y. You can then do a Fit Y by X, with the Group (Ordinal Y) as X and the prediction formula as Y and then do all sorts of tests as you might do with an ANOVA.

Hope this helps,

DS