[Bug]: Logistic regression using GLM no longer works

[patc3]

JASP Version

0.95.1

Commit ID

No response

JASP Module

Regression

What analysis are you seeing the problem on?

Generalized Linear Model

What OS are you seeing the problem on?

Windows 11

Bug Description

I think it's become impossible to run logistic regression in JASP 0.95.1 using the Generalized Linear Model analysis. See report here https://forum.cogsci.nl/discussion/9939/jasp-0-95

If you use a nominal variable that only has text as values+labels, you get this error:

If you try to fix it by adding numeric values, you get this other error instead:

Heart attack.csv

Expected Behaviour

Doing logistic regression with a binary outcome using the Generalized Linear Model analysis in Regression module should work. This works fine in Regression > Logistic Regression.

Steps to Reproduce

1. nominal variable with 2 values (text or values, neither works)
2. run logistic regression using Regression > Generalized Linear Model (the default analysis)
3. try to use that nominal variable as dependent variable

Log (if any)

No response

More Debug Information

-------- Application Info --------
JASP Version: JASP 0.95.1
Build Branch: development
Build Date: Aug 20 2025 13:37:25 (Netherlands)
Last Commit: 180d14c18d13c107ee7c16b0993ff7e1cc02680b

-------- Basic Info --------
Operating System: Windows 11 Version 24H2
Product Version: 11
Kernel Type: winnt
Kernel Version: 10.0.26100
Architecture: x86_64
Install Path: C:/Program Files (x86)/JASP
Platfotm Name: windows
System Local: fr_CA

-------- Extra Info --------
Current code page
Page de codes active�: 850
Page de codes active�: 65001

Final Checklist

• I have included a screenshot showcasing the issue, if possible.
• I have included a JASP file (zipped) or data file that causes the crash/bug, if applicable.
• I have accurately described the bug, and steps to reproduce it.

[boutinb]

The difference with 0.19.3 is that the Dependent Variable in 0.95.1 is set to be a 'scale' variable. But apparently with a Bernouilli type, it expects a nominal. I see in https://github.com/jasp-stats/jaspRegression/blob/master/inst/qml/common/GlmInputComponent.qml that the type of dependent variable was changed by @FBartos in jasp-stats/jaspRegression#403

[FBartos]

The difference with 0.19.3 is that the Dependent Variable in 0.95.1 is set to be a 'scale' variable. But apparently with a Bernouilli type, it expects a nominal. I see in https://github.com/jasp-stats/jaspRegression/blob/master/inst/qml/common/GlmInputComponent.qml that the type of dependent variable was changed by @FBartos in jasp-stats/jaspRegression#403

I think you tagged the wrong person Bruno :) but I created a pull with fix

[JohnnyDoorn]

@boutinb Frantisek made a lovely fix for this; I did notice that you can sort of "hack" the type by specifying a variable for the dependent box, which is restricted to a certain type (e.g., nominal for Bernoulli), and then when you change the family (for instance to "Gaussian"), the type restriction is not re-evaluated and so I could still run a Gaussian model with a nominal dependent variable, even though it should be forced to be a scale variable. Should I make a separate issue for this? I don't think it has the same severity as the current issue, so can maybe wait for the next release.

[boutinb]

@boutinb Frantisek made a lovely fix for this; I did notice that you can sort of "hack" the type by specifying a variable for the dependent box, which is restricted to a certain type (e.g., nominal for Bernoulli), and then when you change the family (for instance to "Gaussian"), the type restriction is not re-evaluated and so I could still run a Gaussian model with a nominal dependent variable, even though it should be forced to be a scale variable. Should I make a separate issue for this? I don't think it has the same severity as the current issue, so can maybe wait for the next release.

Yes, the best is to make another issue for this.