User Community
cancel
Turn on suggestions
Auto-suggest helps you quickly narrow down your search results by suggesting possible matches as you type.
Showing results for 
Show  only  | Search instead for 
Did you mean: 
  • Don’t miss your chance to experience Discovery Summit Europe at our best available rate! Early bird registration through 31 Oct.
  • Graph Builder Toolbar streamlines interactive graphing. Creates shortcuts on Report and GraphBuilder Helpers toolbars. Download and install the extension.
  • JMP will suspend normal business operations for our Rest and Recharge Day on Friday, October 2, 2026.
    Regular business hours will resume on Monday, October 5, 2026.

Discussions

Solve problems, and share tips and tricks with other JMP users.
Choose Language Hide Translation Bar
NoScoped
NoScoped
Level I

When is it appropriate to apply Firth Adjusted Maximum Likelihood and FDR?

Aug 25, 2018 05:36 PM (5070 views)

Currently running a Poisson model in the Generalized Linear Model personality. The personality provides the opportunity to provide Firth Bias-Adjusted Estimates. Additionally, once the model has been run, I have the option to apply a False Discovery Rate to each model effect. Just was curious...when should I use these options? 

0 Kudos
1 REPLY 1
Mark_Bailey
Staff Mark_Bailey
Staff

Re: When is it appropriate to apply Firth Adjusted Maximum Likelihood and FDR?

Aug 27, 2018 09:03 AM (5030 views)  |  Posted in reply to message from NoScoped 08-25-2018

First of all, see Help > Books > Fitting Linear Models. There is a lot of information about fitting and interpreting the GLMs.

 

The Firth bias adjustment has two main purposes. The first purpose is to decrease bias and standard error of the parameter estimate by shrinking the estimate towards zero. This problem is a concern with small data sets or where a predictor is associated with one level in the case of a binomial GLM (e.g., logistic regression). The second purpose of the method is to solve the problem or complete or quasi-separate in the case of logistic regression.

 

The FDR is useful when you have very many simultaneous t-tests (i.e., significant estimate).

Recommended Articles