search

inovex / justcause

💊 Comparing causality methods in a fair and just way.
https://justcause.readthedocs.io/
MIT License
138 stars 15 forks source link

Blog post notebooks and semantic fixes #22

Closed MaximilianFranz closed 4 years ago

MaximilianFranz commented 4 years ago

Through more extensive use in a practical evaluation, I uncovered two major issues, that should not have come through in the first place:

  1. Wrong covariate was used for defining the treatment effects in generators.ihdp
  2. Due to a type in evaluation.py both train and test results were actually "train", thus always yielding the exact same score.

Both are fixed.

FlorianWilhelm commented 4 years ago

lgtm

FlorianWilhelm commented 4 years ago

We could make a new bugfix release after the __future__ annotations problem is fixed. #21