add influence function calculation #37
Conversation
mbannick
left a comment
mbannick
left a comment
There was a problem hiding this comment.
Hi @dbackenroth, thanks for submitting this PR! Could you please add some brief documentation for this in the function?
|
Hi @mbannick I added in some documentation, please let me know what else I can provide. |
R/adjust-logrank.R
Outdated
| se <- sqrt(var_CSL) | ||
| statistic <- U_CSL / se | ||
|
|
||
| if ("id" %in% colnames(df)) { |
There was a problem hiding this comment.
@dbackenroth Thanks for adding documentation explaining this PR!
If someone has a covariate that is named id (that would be a strange name for a covariate, but it's possible, particularly if they are doing stratification by some type of id), then this may produce unintended behavior. I don't want anything to break for current users since this is a stable package.
Could you make the following change please:
- Modify
robincar_logrankso that it takes areturn_influenceargument that is by default FALSE, and then propagate that argument tomodeland then it can be accessed here and below, asif model$return_influencerather thanif "id" %in% ...? This way, users won't see any change in behavior based on default usage.
There was a problem hiding this comment.
@dbackenroth it also looks like some tests fail with the current PR. Could you please ensure the test suite runs through successfully? Thanks! See e.g., https://github.com/mbannick/RobinCar/actions/runs/20396466576/job/58744492353?pr=37.
There was a problem hiding this comment.
@mbannick hi, i added in unit tests. two of the tests are failing for me but I don't think that has to do with my code (test-glm-calibration.R: test calibration and test-glm-legacy.R: GLM legacy), it's related to functions not being found, e.g., str_detect. I added in two columns to robincar_tte signature, one is return_influence and the other is id_col. i may have gone overboard with the error checking, please let me know if you'd like me to make any further changes
2 participants
calculates influence function for CALRT and returns to user (if user includes an "id" column in input dataset)