Conversation
remlapmot
left a comment
remlapmot
left a comment
There was a problem hiding this comment.
Very nice Ryan
I have run the tests locally under Python 3.10, 3.11, 3.12, 3.13, and 3.14 and all pass.
|
I think let's leave the GitHub Actions tests workflow for now - hopefully GitHub will fix whatever is going on with their runners once folk return after the break. |
When: * treatment_level=[0, 1] (binary control vs treatment) * method="censoring" * weighted=True
|
In converting our short course practical from R to Python I found a tiny bug related to needing to use I have implemented this in the commit I've added to this branch (I have checked all tests pass locally under Python 3.10 and 3.14) - if you prefer I could remove it and send as a separate PR after you've merged this one (but given it's just you and me atm, hopefully ok to do this). |
remlapmot
force-pushed
the
compevent-fix
branch
from
a842eb4 to
90dca14
Compare
|
Maybe I should move my commits to another branch - I'm having all sorts of trouble with converting our practical - I think the problem is with the categorical variables. |
|
Also in the bootstrapping I am getting alot of problems with perfect separation, which somehow we didn't seem to encounter in R. |
|
(Even with these commits our example still fails -- will continue tomorrow or at w/e or next week) |
|
At least in it's current iteration, seems to be only 3.14 which consistently fails.
|
2 participants
Despite the mostly misnamed branch, this PR offloads models out of memory once they are built.
Additionally, in my testing, I found a bug with pre-expansion bootstrapped models - this has been fixed.
I've noticed that the tests have been failing due to runner failure. I'm unsure if this is repeatedly on GH's end (since this started as they've changed runners), or if we should change/skip the test.