R/estim_CS.R

In MatthieuStigler/multiDiff: Multi-year diff-diff and FE decompositions

utl_char_to_num <- function(x) match(x, unique(x))

if(FALSE){
  utl_char_to_num(rep(LETTERS[1:5], each=3))
}


#' Apply Callaway and Sant'Anna (2021) estimate
#'
#' Simple wrapper to function \code{\link[did]{att_gt}} from package did.
#' @template param_mdd_dat
#' @param ... further objects passed to \code{\link[did]{att_gt}}
#' @returns Output of \code{\link[did]{att_gt}}, therefore of class \code{\link[did]{MP}}.
#' @examples
#' dat <- sim_dat_staggered(as_mdd = TRUE, Time=6)
#' mdd_CS(dat)
#'
#' ## with a single treatment time, this is equivalent to mdd_event_study()
#' dat_common <- sim_dat_common(as_mdd = TRUE, timing_treatment = 5:10)
#' all.equal(unname(coef(mdd_event_study(dat_common))),
#'           broom::tidy(mdd_CS(dat_common, base_period = "universal"))$estimate[-4])
#' @export
mdd_CS <- function(mdd_dat, ...){

  requireNamespace("did")
  vars <- intrnl_mdd_get_mdd_slot(mdd_dat)$var_names

  ## conv to did format
  dat_did <- mdd_conv_mdd_to_did(mdd_dat)
    # mutate(treat_timing2= if_else(.data$treat_timing>0, min(.[[vars$time.index]])+.data$treat_timing, 0))

  ## make unit.index (idname) numeric
  if(!is.numeric(dat_did[[vars$unit.index]])){
    nam <- vars$unit.index
    dat_did <- dat_did %>%
      mutate({{nam}} := utl_char_to_num(vars$unit.index))
  }

  ## now run
  did::att_gt(yname = vars$y_var, tname = vars$time.index, idname = vars$unit.index,
              gname = "treat_timing",
              data = dat_did,
              ...)
}

if(FALSE){
  dat <- sim_dat_staggered(as_mdd = TRUE, Time=6)
  dat
  environment(mdd_CS) <- environment(mdd_data_format)
  mdd_CS(dat)

}