sample_splitting_estimation: sample_splitting_estimation

In yasui-salmon/stRatification: estimate the treatment effect in subgroup of dataset

Description

Estimate average treatment effect in each subgroup created from predictionresult.

Usage

sample_splitting_estimation(data, Y, treatment, X, M, ntilen, adjusted)

Arguments

data	an optional data frame containing the variables in the model.
Y	string which indicates the outcome variable in the estimation.
treatment	string which indicates the treatment variable.
X	string that indicates the control variables. you are able to set like c("a", "b") to implement multiple control variables.
M	numeric that how many time will you iterate the splitting estimation. M in Abadie et al. (2016).
ntilen	int that show how many group will you make in the dataset
adjusted	TRUE/FALSE to set which estimation you want to implement, adjusted estimation or unadjusted estimation.

Examples

##---- Should be DIRECTLY executable !! ----
##-- ==>  Define data, use random,
##--	or do  help(data=index)  for the standard data sets.

## The function is currently defined as
function (data, Y, treatment, X, M, ntilen, adjusted)
{
    control_set <- data %>% filter_(paste(treatment, "==", 0))
    treatment_set <- data %>% filter_(paste(treatment, "==",
        1))
    factor_col <- which(sapply(data, class) %in% c("character",
        "factor"))
    M <- M
    result <- data.frame()
    for (i in 1:M) {
        sample_flag_control <- sample(NROW(control_set), NROW(control_set)/2,
            replace = F)
        prediction_set <- control_set[sample_flag_control, ]
        estimation_set <- control_set[-sample_flag_control, ]
        estimation_set_cut <- level_cut(dataset = estimation_set,
            prediction_set = prediction_set, factor_col = factor_col)
        treatment_set_cut <- level_cut(dataset = treatment_set,
            prediction_set = prediction_set, factor_col = factor_col)
        ntile_formula <- as.formula(paste(Y, "~", paste(X, collapse = " + ")))
        reg <- prediction_set %>% lm(data = ., formula = ntile_formula)
        out <- ATE_ntile(model = reg, Y = Y, X = X, treatment = treatment,
            adjusted = adjusted, estimation_dataset = estimation_set_cut,
            treatment_dataset = treatment_set_cut, ntilen = ntilen,
            i = i)
        result <- rbind(result, out)
    }
    return(result)
  }

yasui-salmon/stRatification documentation built on May 4, 2019, 2:31 p.m.