compute F(y|ytmin1) where F is the conditional distribution of untreated potential outcomes for the treated group conditional on ytmin1. This is computed under the copula stability assumption. This function is typically computed internally in the csabounds package but is provided here for convenience.

F.Y0(
  firststep = c("dr", "qr", "ll"),
  xformla,
  data,
  yvals,
  tvals,
  h = NULL,
  retF = TRUE,
  retZ = FALSE
)

Arguments

firststep

whether to use distribution regression ("dr"), quantile regression ("qr"), or local linear distribution regression ("ll") for the first step estimation of condtional distributions

xformla

a formula for which covariates to use

data

the data.frame. It should contain colums called "Y1t" and "Y0tmin1" which correspond to treated potential outcomes in the third period and untreated potential outcomes in the second period

yvals

Sequence of values to compute distributions over (currently not used as these are determined internally)

tvals

Sequence of values to compute conditional distribution over (currently not used as these are determined internally)

h

optional bandwidth when using local linear regression

retF

whether or not to return the distribution function itself; the default is TRUE. If false, it returns a distreg::DR object

retZ

whether or not to return the distribution of the transformed random variables due to the copula stability assumption. This is mainly used in the numerical bootstrap procedure and, therefore, the default is FALSE.

Value

distribution F(y|ytmin1)