GroupIPW: Estimating the group average potential outcome

View source: R/GroupIPW_function.R

GroupIPWR Documentation

Estimating the group average potential outcome

Description

IPW estimator of the group average potential outcome.

Usage

GroupIPW(
  dta,
  cov_cols,
  phi_hat,
  gamma_numer = NULL,
  alpha,
  neigh_ind = NULL,
  trt_col = NULL,
  out_col = NULL,
  alpha_re_bound = 10,
  integral_bound = 10,
  keep_re_alpha = FALSE,
  estimand = c("1", "2"),
  verbose = TRUE
)

Arguments

dta

Data frame including treatment, outcome and covariates.

cov_cols

The indices including the covariates of the ps model.

phi_hat

A list with two elements. The first one is a vector of coefficients of the ps, and the second one is the random effect variance. If the second element is 0, the propensity score excludes random effects.

gamma_numer

The coefficients of the ps model in the numerator. If left NULL and estimand is 1, the coefficients in phi_hat will be used instead.

alpha

The values of alpha for which we want to estimate the group average potential outcome.

neigh_ind

List. i^th element is a vector with the row indices of dta that are in cluster i. Can be left NULL.

trt_col

If the treatment is not named 'A' in dta, specify the treatment column index.

out_col

If the outcome is not named 'Y', specify the outcome column index.

alpha_re_bound

The lower and upper end of the values for bi we will look at. Defaults to 10, meaning we will look between - 10 and 10.

integral_bound

The number of standard deviations of the random effect that will be used as the lower and upper limit.

keep_re_alpha

Logical. If set to TRUE the "random" effect that makes the average probability of treatment equal to alpha will be returned along with the estimated group average potential outcome.

estimand

Character, either '1' or '2.' If 1 is specified, then the estimand with numerator depending on covariates is estimated. If estimand is set equal to 2, the numerator considered is the product of Bernoulli.

verbose

Whether printing of progress is wanted. Defaults to TRUE.


gpapadog/Interference documentation built on Oct. 30, 2022, 9:20 p.m.