LendieFollett/A-New-Kind-of-Impact-Evaluation

Bayesian framework for estimating overall capabilities as well as treatment effects on capabilities, choices, and functionings in the Kenya's unconditional cash-transfer scheme, Cash Transfer for Orphans and Vulnerable Children (CT-OVC) README ================ # A New Kind of Impact Evaluation Bayesian framework for estimating overall capabilities as well as treatment effects on capabilities, choices, and functionings in the Kenya’s unconditional cash-transfer scheme, Cash Transfer for Orphans and Vulnerable Children (CT-OVC). ## Usage Within your project location, save functions.R, analysis.R, BSFA_model.stan, in subdirectory “src” and ctovc_final.csv in subdirectory “raw_data”. Set your working directory to be your project location You can use setwd(). For example, ``` r setwd("/Users/Follett/Documents/Research/BSFA") ``` Within this directory, create two new folders named "Original" and "Matched". This is where the corresponding results and plots will be saved. Before you can load the packages, you will need to install them if you have not already done so. Code may look like: ``` r install.packages("rstan") #MCMC sampling install.packages("tidyverse") #graphics, piping, summaries, data manipulation install.packages("reshape2") #wide-to-long melting install.packages("tidyr") #data manipulation install.packages("randomForest") #estimate propensity scores install.packages("Matching") #matching treatment, control groups ``` After each package is installed you can proceed with the code in analysis.R starting with loading each of the above packages. - anthro.do includes the code that is used to compute caloric variables. - analysis.R will source functions.R, which contains functions for stan sampling, treatment effect estimation, and plot creation. Model estimation begins with code chunk that looks like ``` r diversity_sampled <- do_sampling(y=y_stand(kenya$diversity, kenya), X=full_X, X_cntr = full_X_cntr, hh_id=kenya$hhcode, loc_id=kenya$location, file = "src/selection_model2.stan", kappa = 1) ``` Running this should result in 4 chains being distributed across 4 cores: The rest of the code will fit the models for the remaining responses and create graphics and tables present in the …