Closed damonbayer closed 1 week ago
Attention: Patch coverage is 95.61404%
with 5 lines
in your changes missing coverage. Please review.
Project coverage is 92.39%. Comparing base (
f68bea6
) to head (5acbd0e
). Report is 10 commits behind head on main.
:umbrella: View full report in Codecov by Sentry.
:loudspeaker: Have feedback on the report? Share it here.
I like getting lighter code. I would just wait for @dylanhmorris' take on this before merging.
Noting that I updated the original description for anyone who had previously read it.
Once this is merged I think the
_rv
suffix for those class attributeRandomVariables
can be removed, but that's a separate PR.
Happy to discuss further. I know people are generally opposed to variable names that state the variable type, but I think it is an easy remedy to avoid having code like latent_hosp_admissions = latent_hosp_admissions.sample
.
I have mangled the git history here somehow. Retrying on https://github.com/CDCgov/multisignal-epi-inference/pull/207.
I don't see any reason to have functions in the class that just call one line functions. We might as well call them directly. We actually do this already for some parts of the model, so this is also a more consistent approach.
My overall ambition is to make models as lightweight as possible and encourage users to make models with minimal overhead. Possibly, going as far as to not provide any prebuilt models in the package itself (but provide ample examples in documentation).