Incremental propensity score intervention

Description

Creates an incremental propensity score intervention (IPSI) that multiplies each individual’s odds of treatment by delta. Values of delta > 1 increase the probability of treatment; delta < 1 decrease it. This is a stochastic modified treatment policy indexed by a single scalar, and corresponds to the "shift-the-odds" counterfactual studied by Kennedy (2019).

IPSI is binary-only and is only meaningful under estimator = “ipw” – the IPW engine computes the Kennedy closed-form weight \(w_i = (\delta A_i + (1 - A_i)) / (\delta p_i + (1 - p_i))\) directly from the fitted propensity, with no counterfactual treatment value to predict at. estimator = “gcomp” and estimator = “matching” therefore do not support ipsi().

Usage

ipsi(delta)

Arguments

delta Positive numeric. The odds multiplier.

Value

A causatr_intervention object.

References

Kennedy EH (2019). Nonparametric causal effects based on incremental propensity score interventions. Journal of the American Statistical Association 114:645-656.

See Also

static(), shift(), dynamic(), scale_by(), threshold()

Examples

library("causatr")

fit <- causat(nhefs, outcome = "wt82_71", treatment = "qsmk",
              confounders = ~ sex + age + wt71)
contrast(fit, interventions = list(
  double_odds = ipsi(2),
  half_odds   = ipsi(0.5)
))