Input check

[judithabk6]

• addresses Failing estimators with input shape (n,1) #75 implement an input checker for all estimators #57 need to correct IPW code for y and t input format #14 m.astype(int) for multiply_robust #41
• implementation of an input checker for all estimators
• adding ravel of m when needed
• removes the xfail of the test_robustness_to_ravel format (

  med_bench/src/tests/estimation/test_get_estimation.py

  Line 257 in 4ebee9c

  def test_robustness_to_ravel_format(data, estimator, config, effects_chap):

  )

still a draft (I know there are commented lines to remove or this sort of things), I have issues with

• exact tests, not sure why
• maybe it is the occasion to deal with the reshapes in the src/med_bench/utils/nuisances.py file (need some input on how it should be done @houssamzenati)

I open the PR to discuss if the main direction is ok for everyone (sorry, in develop, not main)

[bthirion]

Looks good.

[houssamzenati]

The changes look good to me. Let me know if I can suggest on nuisances.py with the changes that I can state here again:

delete lines 122-125
if len(x.shape) == 1:
x = x.reshape(-1, 1)
if len(m.shape) == 1:
m = m.reshape(-1, 1)

delete line 167-168
if len(x.shape) == 1:
x = x.reshape(-1, 1)

• replace line 231-236:
  if len(x.shape) == 1:
  x = x.reshape(-1, 1)
  if len(m.shape) == 1:
  mr = m.reshape(-1, 1)
  else:
  mr = np.copy(m)
  by mr = np.copy(m)

[bthirion]

The PR looks good, but probably needs more tests.

[judithabk6]

So tests pass, however I think they should not. I missed this modification of tests, 4f63b80 but it results on the exactness tests being no super useful as they do not catch modifications of the behavior of the estimators. And the threshold of 0.01 is larger than the indirect effect so it's really permissive

[codecov-commenter]

⚠️ Please install the to ensure uploads and comments are reliably processed by Codecov.

Codecov Report

Attention: Patch coverage is 97.43590% with 1 line in your changes missing coverage. Please review.

Project coverage is 89.08%. Comparing base (43c118b) to head (7a08ff8).

Files	Patch %	Lines
src/med_bench/mediation.py	90.90%	1 Missing ⚠️

❗ Your organization needs to install the Codecov GitHub app to enable full functionality.

Additional details and impacted files

@@             Coverage Diff             @@
##           develop      #81      +/-   ##
===========================================
+ Coverage    86.53%   89.08%   +2.55%     
===========================================
  Files            6        6              
  Lines          750      733      -17     
===========================================
+ Hits           649      653       +4     
+ Misses         101       80      -21     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[judithabk6]

The PR looks good, but probably needs more tests.

The PR looks good, but probably needs more tests.

indeed. I think it would be nice to test each case where the check_input function raises + the output format. I don't know if you have other tests in mind @bthirion ?

As discussed @brash6 we can work on it together depending on who has time first

[bthirion]

Sorry, I would inspect uncovered lines of code.

[bthirion]

The changes LGTM. But we need a green CI...

[judithabk6]

I have modified the tests to make them simpler and to split them (instead of testing several things in the same test). Not sure which one is best. Let me know, but otherwise, I think it's ok, and it will allow to merge the fixes of the CI.

[bthirion]

LGTM, besides one small detail.

[bthirion]

If you can do a pass to have consistent argument order, that would be great.

[bthirion]

Please make sure that the argument order y, t, m, x is there in all functions. Sometimes, there is t, m, x, y

[bthirion]

LGTM. Will push 2 minor things