This repo is currently hosted in gitlab. Click here.

Gretl-logit-test

Script for the Data Science A-Z by SuperDataScience course. Allows for very simple model testing without the limitations of the point & click interface of Gretl.

Keep in mind that it has a questionable quality, but it does the job.

Thanks to the Gretl user list guys for helping me too :)

How-to

The script is pretty much self explanatory. Download the .INP files (csvs should already be provided, but the ones in the repo are valid too), double click applying-model-to-test-data.inp, and follow the instructions contained in the comments.

It has been tested in Windows and Linux, with Gretl 2019a

Output

The script saves two files.

• training-output.csv -> Contains RowNumber, Exited and the model probability scores
• forecast-output.csv -> Contains RowNumber and the testing probability scores

References for learning about Gretl scripting

Gretl has not been subject to much attention outside academia, I guess because the documentation is a bit lacking, with mostly PDFs in the wild. It is a nice little self-contained tool though.

If you're looking for documentation:

• For the total beginner: The Gretl guide (PDF)

• Assuming you know a bit of programming already: Hansl Primer (PDF)

• HTML references: Command Reference & Function Reference