Minor changes to tests to anticipate breaking changes in the upcoming marginaleffects.
Background: I've had to deal with several bugs related to the order of rows in marginaleffects output. In previous versions, I was trying to be clever to display rows "nicely" based on by and other arguments. Unfortunately, it turns out to be very tricky to handle the interplay of all the options: by, hypothesis, posterior_draws, etc.
I no longer trust the sorting code and will thus release a new version of marginaleffects which does not sort automatically. The breaking change is that the rows in the output will not always be the same as in previous versions.
This PR just adds a sort call to your tests so that comparisons match.
Not sure if you still run the tests on CRAN, but if so you'll have to make a new release. Sorry!
grantmcdermott
commented
1 year ago
Minor changes to tests to anticipate breaking changes in the upcoming
marginaleffects.Background: I've had to deal with several bugs related to the order of rows in
marginaleffectsoutput. In previous versions, I was trying to be clever to display rows "nicely" based onbyand other arguments. Unfortunately, it turns out to be very tricky to handle the interplay of all the options:by,hypothesis,posterior_draws, etc.I no longer trust the sorting code and will thus release a new version of
marginaleffectswhich does not sort automatically. The breaking change is that the rows in the output will not always be the same as in previous versions.This PR just adds a sort call to your tests so that comparisons match.
Not sure if you still run the tests on CRAN, but if so you'll have to make a new release. Sorry!