Subscribe Bookmark RSS Feed

Selecting Beta distribution for a Generalized Linear Model personality in JMP 12 or 13

SoilSAS

Community Trekker

Joined:

Nov 17, 2016

I have many response variables (composition/abundance data) that have beta distributions and despite the fact that JMP 12.1.0 Help tells me "The Generalized Regression personality enables you to specify a variety of distributions for your response variables. The distributions fit include normal, Cauchy, exponential, gamma, Beta, binomial..." I can't select a Beta distribution when the GLM personality is slected. I only have 4 options I can choose from: normal, binomial, Piosson, and exponential. Does anyone know how to perform a GLM (mixed) for Beta distributed data in JMP 12? I also have a license for JMP 13 that I haven't yet installed and so if anyone know how to do that in 13, I would also be grateful. Thanks in advance.

2 REPLIES
SoilSAS

Community Trekker

Joined:

Nov 17, 2016

Another question: If I define my response variables as having a Beta distribution (via column properties), then should I select "Normal" for distribution and then select a logit link function? (since Beta distributions should be run with a logit link function in GLMs)
Peter_Bartell

Joined:

Jun 5, 2014

I think you might be confusing the Generalized Regression subplatform with the Generalized Linear Model subplatform within the Fit Model personality dropdown menus. The beta distribution, as stated in the JMP documentation, is only supported in the Generalized Regression subplatform. Generalized Linear Model does not support that response functional form. And Generalized Regression is only supported in JMP Pro...your original post states only "JMP"...so you won't have access to the Generalized Regression subplatform if you are running JMP and not JMP Pro.