Closed nilshg closed 8 months ago
I should add I see a failure on nightly but that was there before these changes so warrants a separate investigation.
All modified and coverable lines are covered by tests :white_check_mark:
Comparison is base (
8c99175
) 91.24% compared to head (31dac02
) 91.18%.
:umbrella: View full report in Codecov by Sentry.
:loudspeaker: Have feedback on the report? Share it here.
Thanks, looks good. For some reason the estimates are slightly different on Julia nightly, not sure what's going on (but agree that this is orthogonal to the PR).
This should fix #54
I'm not sure that this is correct, but I've basically changed
VcovDataGLM
in the same way thatVcov
changed in the upstream package, and made the relevant changes to functions that use this. I'm not sure I understood the dependency structure betweenVcov
,StatsAPI
andGLFixedEffectModels
so do check if this makes sense.Unrelated changes:
cholesky!(..., Val(true))
but had to make that conditional on version as this was only introduced in 1.7.x
variable local in thenlreg
tests at the end to get rid of the soft scope warnings.