Truncate and Calibrate Propensity Scores — truncate

The truncation_method parameter controls how the propensity scores are calibrated and truncated. - "isotonic": Performs isotonic calibration, providing both data-adaptive truncation and calibration for the propensity scores. - "adaptive": Adapts the truncation level using a loss function for the inverse propensity score. - "deterministic": Bounds the estimates away from 0 and 1 using the threshold 25/sqrt(n)/log(n). - "none": Bounds the estimates away from 0 and 1 using the threshold 1/n.

Usage

truncate_propensity(
  pi.hat,
  A,
  treatment_level = max(A),
  truncation_method = c("isotonic", "adaptive", "deterministic", "none")
)

Arguments

pi.hat: A numeric vector containing estimates of the propensity score for treatment_level.
A: A numeric vector of treatment values.
treatment_level: A numeric value indicating the treatment level to calibrate.
truncation_method: A character string indicating the truncation method: "isotonic", "adaptive", "deterministic", or "none".

Value

A numeric vector of truncated and calibrated propensity scores.

Details

This function truncates and calibrates propensity scores to ensure bounded values and improve their reliability.