search

alexanderquispe / 14.38_Causal_ML

JupyterNotebook for the MIT course
MIT License
13 stars 5 forks source link

Notebook_Inference_clustering #47

Open anzonyquispe opened 3 years ago

anzonyquispe commented 3 years ago

We have to replicate this kaggle notebook.

anzonyquispe commented 3 years ago

Laso estimations

When We used the Sklearn Lasso function, We got similar results in Python and R for the argument post = FALSE. However, we got different results for post = TRUE, but they are not far between themselves. Additionally, we added regressions with hdmpy library to inspect results. It returns better results for post = TRUE.

cv.glmnet curiosities

According to link1 and link2, It standardize X variables before estimation. In sklearn we have an option named normalize. It normalizes X by subtracting the mean and dividing by the l2-norm. Based on link3, cv.glmnet (standardize = TRUE) and sklearn (normlize = True) are not the same.

Random Forest curiosities

We added some extra hyperparameters to RF models to be equal to the R default hyperparameters.