Dear gretl developers,
When I am teaching ordinary least squares, I think that it is important that the
normality assumption should be tested. This is easily done in gretl with the
nice normality test that is easily accessible after running the OLS regression
model.
When the normality assumption is rejected, I would like to suggest to my student
to use bootstrap to calculate confidence intervals and p-values instead.
However, most students have no experience of programming at all, which make it
hard for them to perform the bootstrap calculations on their own. They
therefore need an easy to use point-and-click GUI to be able to perform the
necessary bootstrap calculations.
My suggestion is that in the OLS output window a new menu entry Analysis ->
BootstrapÂ… is included, where the user can choose to perform bootstrap
calculations for the model by clicking on check boxes, radio buttons, etc.
I think that the user should have access to the following bootstrap choices:
* Choice between residual bootstrap (resampling the residuals) and paired
bootstrap (resampling the pair y,x)
* Choice of number of bootstrap resamples
* Choice of calculating p-values for each explanatory variable
* Choice of calculating 95% confidence intervals for each explanatory variable
Easily accessible bootstrap calculations are (at least to my knowledge) not
accessible form other major GUI statistical soft wares like SPSS, MINITAB,
etc., so I think that this should be very useful for many persons and be an
incentive to using gretl. It could be close to a killer application.
Thanks for great software, and I hope you will consider my suggestion.
Best regards,
Andreas Karlsson