Date: 30.08.2022
This repository contains the R code to implement the direct M-out-of-N bootstrap for nearest neighbor matching along with the code used in the simulation study in:
Walsh, C.P. and Jentsch, C. (2022+), "Nearest neighbor matching: Does direct M-out-of-N bootstrapping without bias correction work when the naive bootstrap fails"
This repository contains:
- R code in the file moon_matching.r to implement our M-out-of-N bootstrap. Further details are given in the file.
- R code in the file simulation_function.r to do one simulation run for a fixed ratio of treated to control units over various sample sizes and bootstrap resample sizes. Further details are given in the file.
- R code used in the simulation study of the paper for the ratio of treated to control given by alpha = 0.2, 1, 2.84 and 5 is in the files main_file_alpha_02.r, main_file_alpha_1.r, main_file_alpha_bar.r, main_file_alpha_5.r respectively.
- the estimation results from the simulation study are contained in the folder Results/ in the files ATET_MooN_alpha_0.2.RData, ATET_MooN_alpha_1.RData, ATET_MooN_alpha_2.84.RData and ATET_MooN_alpha_5.RData
- the folder Results/ also contains the file summarize_results_paper.r to construct:
- paper_plot_point_estimates.pdf, which corresponds to Figrue 1 of the paper
- and Coverage_Table.tex, which gives the coverage probabilities in Table 1 of the paper.
ACKNOWLEDGEMENTS: The simulations were run on the supercomputer CLAIX at RWTH Aachen University as part of the NHR4CES infrastructure with computing resources under the project p0020105.
CAUTION: The simulations in the files main_file_alpha_02.r, main_file_alpha_1.r, main_file_alpha_bar.r, main_file_alpha_5.r were run on 48 cores with run times in hours of 07:43:36, 14:00:32, 10:59:47 and 07:56:22 respectively.