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; } } }

Generalized Regression/Linear Model with Random Effects

Peter_Polito

Since beginning at JMP I have had multiple instances of a customer/potential customers frustrated by the inability to run a Generalized Linear model with Random Effects. If you have not attempted this yourself, you will be allowed to select the Generalized Linear Model personality (or Generalized Regression when I've demo'd Pro), construct the model, and even denote which variables are random, but when you click 'run,' you receive a message that states, "Random Effects are not supported by Generalized ... " I don't have a lot of data, but I'm averaging a request for this feature once every ~2-3 months, which seems high. Am I an outlier or is this a regularly requested feature in other territories?

Unfortunately, this happens most often with prospective customers and so motivating them to post to the public wish list has not been terribly successful, thus I am posting it here.

[editor's note]

Here's the public version of this wish:

Generalized linear multilevel modeling

Status: In Review

Comments

mia_stephens

Thanks for posting this Peter. There are many requests from academics for Generalized Linear Mixed Models (i.e. mixed models with categorical responses, logistic regression with random effects,...) - I think this came up at least once a month when I was an Academic Ambassador.

We'd be curious in understanding how often commercial customers are requesting this functionality. If you're fielding requests for GLMM, please kudo this post and share your experiences in the comments: The industry, the customer (if you can), the types of problems the customer is trying to solve that would require this, and what they're currently using to analyze these data (R, SAS, other,...?).

Thanks!

Peter_Polito

Most recently this came up during a Pro demo. (Potential) customer is QTS (https://www.qtsdatacenters.com/), they did not mention why they needed the feature nor what method they are currently using. This previously came with a few other prospectives, Spectral MD (https://www.spectralmd.com/, requested by a biostatistician) being the one that comes to mind at the moment.

Jeff_Perkinson

Looking at the public version of this wish I see kudos from WL Gore, P&G, ST Micro, Endress+Hauser, Elanco and more.

ruthhummel

I hear this request often from academics. One requester is the agriculture / plant breeding program at North Dakota State University. They presented at JMP Discovery (US) last year and are presenting at an ag conference in November to show a JMP add-in they created to allow breeders to enter data in the field and have this data be pulled into tables and cross-tabs and then used in a mixed model to estimate BLUPs. However, JMP does not allow the models they need, so they actually then use ASREML as their modeling platform. They would love to be able to do this in JMP instead.

Jeff_Perkinson

Here's a few other posts in the community around this idea:

Fitting generalized linear mixed models

Generalized Linear Mixed Model (GLMM) JMP Pro 13 ?

Generalized linear models

How do I set up a Generalized Linear Mixed Model in JMP 9?

PROC GLIMMIX capabilities implementation in the Generalized Regression personality (Dow Wish List)

Mixed models >> categorical data >> random effect functionality (P&G Wish List)

mia_stephens