Skip to contents

Theorem 5.2: the variance-minimising trimming threshold

Source: R/tmlefp_native.R
morie_optimal_alpha.Rd

Solves gamma = 2 E[k(X) | k(X) < gamma] for gamma by fixed point iteration. Homoskedastic unless conditional variances are supplied.

Usage

morie_optimal_alpha(
  pscore,
  sigma2_treated = NULL,
  sigma2_control = NULL,
  tol = 1e-12,
  max_iter = 200
)

Arguments

pscore

Numeric vector of propensity scores in (0, 1).

sigma2_treated, sigma2_control

Optional conditional variances.

tol, max_iter

Convergence tolerance and iteration cap.

Value

A list with alpha, gamma, keep, trim, no_trimming, k.

References

Crump, R. K. et al. (2009). Theorem 5.2.

Examples

set.seed(22)
ps <- plogis(rnorm(1000, 0, 2))
morie_optimal_alpha(ps)[c("alpha", "gamma")]
#> $alpha
#> [1] 0.09635781
#> 
#> $gamma
#> [1] 11.48462
#>