CODED using Chernozhukov (2018) paper to print
- Based on Python Codes in below site by Paul Schrimpf, reformulated codes that replicated the results in Chernozhukov(2018) paper. https://datascience.quantecon.org/applications/heterogeneity.html
-The results print:
-
Stage 1: Best Linear Predictor(BLP) of Conditional Average Treatment Effect (CATE) using 4 machine learning techniques -> Elastic Net, Boosted Tree, Neural Network, Random Forest
-
Stage 2: Sorted Group Average Treatment Effects (GATES): 20% most, 20% least affected groups, difference
-
Stage 3: Classification Analysis (CLAN) or average characteristics of the most and least affected groups