Train-test split for logistic regression
Is there a way in JASP to run logistic regression on training data and test it on separate data? I could use a machine learning method to split the data and label the training rows, which would allow fitting of a logistic regression model to only the training data. But I don't know how I could then evaluate the model's predictive performance on test data.
Comments
JASP has a machine learning module that does the splitting, but I don't think we do this for logistic regression. I guess you could do this, but given that all of the standard statistical machinery is already available, I am not sure why you would want to do it. But I'll bring this to the attention of our machine learning experts.
Cheers,
E.J.
Hi Quinnculver,
This is indeed not possible at the moment. If you would like to see this at some point in the future, please create a feature request at https://github.com/jasp-stats/jasp-issues/issues?q=sort%3Aupdated-desc+is%3Aissue+is%3Aopen (see also https://jasp-stats.org/2018/03/29/request-feature-report-bug-jasp/). We currently support "Linear" and "Regularized linear" under "Regression" in the machine learning module and I think it would be sensible to mirror these for classification (e.g., "logistic" and "Regularized logistic").
Cheers,
Don
FYI, looks like this was added since I asked: the Machine Learning module now has "Logistic / Multinomial Regression Classification".
Yes, it was added in 0.19.2 by Koen Derks.