This Github repository contains replication files for Park and Kang (2019).
ECLSK folder contains 1.Analysis.R, and 3.Summary.R files, and replicates the data analysis in Section 6 of the paper.
-
1.Analysis.R file cleans the raw data (available at https://nces.ed.gov/ecls/dataproducts.asp), implements our method, and generates Result_BATCH####.csv files in NData folder.
-
2.Summary.R file summarizes Result_BATCH####.csv files and estimates the groupwise effects based on the causal forest using
grf
package. -
NData folder contains csv files titled as BATCH_BATCH####.csv. Here B#### represents the index of sample splitting procedure ranging from B0001 to B0100.
Simulation folder contains plot, and Result, Summary_Result folders, and 1.Estimation.R and 2.Summary.R files.
-
1.Estimation.R replicates the simulation analysis in Section 4 of the paper. The result of the simulation is saved in Result folder.
-
2.Summary.R summarizes the results in Result folder.
-
Result folder contains csv files titled as Result_###_COR_###_rho###_B####_SubB####.csv. The first ### takes either CPS or VPS according to the propensity score model. The second ### takes either Est or CL according to the i.i.d. model or the clustered model. The third #### varies from 0000 to 0003 according to the effect heterogeneity. The last #### varies from 0000 to 0100 according to the index of sample splitting procedure.
-
Summary_Result folder contains csv files titled as Result_###_COR_###_rho###_B####.csv which is the merged file of the csv files in Result folder with the format Result_###_COR_###_rho###_B####_SubB####.csv.
-
plot folder contains graphical summary of the simulation results.
-
MySL.R contains functions used for implementing superlearner algorithm and estimating the nuisance functions.
-
SSLS.R contains functions used for the estimation of the groupwise effect using the methods in Section 2 and 3 of the main paper.
Chan Park & Hyunseung Kang (2019) A Groupwise Approach for Inferring Heterogeneous Treatment Effects in Causal Inference, arXiv:1908.04427 [link]