Regression: Casewise Diagnostic and wrong case numbers
Hi,
I have a dataset with an ID variable and I calculate a regression (using JASP v0.17). The table for casewise diagnostics (std. residuals) is showing 4 cases with a number. Actually, I would think that these cases are the lines in my dataset which are the same numbers in my id variable. But it is not. Some cases in the JASP table aren't even close to the id variable numbers. So how is JASP counting or how do I get to the correct lines in my dataset, when I wouldn't use a case id. Currently, I would filter wrong cases for my regression.
Thanks
Nick
Comments
Hi Nick
This looks like an issue for our GitHub page. Could you include screenshots that showcase the problem? Perhaps it is a bug and then the programmers would be keen to fix it. For details see https://jasp-stats.org/2018/03/29/request-feature-report-bug-jasp/
Cheers,
E.J.
Unfortunately, it's a dataset with customer data and I can't provide any screenshots. But this problem should be the same with every example dataset I think...
Using the example dataset "Album Sales" for regression, the line number seems to be correct in the casewise diagnostics... This is strange. I would have done a completely wrong diagnosis with my customer data when I wouldn't have double checked the results.
So, I finally found out what the problem is. My depend variable has missing values, so not all of my data is used in the regression. It seems that JASP is numbering the cases in the collinearity regression just by renumbering in background. The frontend user does not have any chance to find out how the new number is connected to its original line in the dataset. This is really a bug.
I'll ask the team whether the latest release has fixed the bug. I recall this also being posted on our GutHub page but I can't find the issue. Thanks for looking into this.
Hi Nick,
This should be fixed in 0.17.1!
Cheers,
Don