Skip to content

Conversation

@patrick-schultz
Copy link
Member

Fixes #8349.

The anyMissing implementation on PArray was wrong, and could return false if the only missing elements were in the last aligned 32 bits containing the end of the missing bits. This in turn let some rows be included in the regression even if some of the covariates were missing.

Copy link
Contributor

@tpoterba tpoterba left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

need a test, however weird+hard to reproduce this bug is.

@patrick-schultz
Copy link
Member Author

It uses hasMissingValues now, which already has a test. I'm beefing up that test now to be safe. test_linreg fails if you swap the intercept with the other covariate, I can add that too.

@tpoterba
Copy link
Contributor

Yeah, a test of linreg is what I meant, thanks.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

error in linreg aggregator based on placement of '1'

3 participants