Value of a treatment regime (AIPW / IPW / regression)

Estimates the policy value of a binary treatment regime and compares it against treat-all and treat-none. near_indifferent is the share of units whose estimated effect lies within a tenth of a standard deviation of zero – a boundary-mass heuristic, not a per-unit standard-error test.

Usage

morie_regime_value(
  y,
  d,
  X,
  regime,
  propensity = NULL,
  method = c("aipw", "ipw", "regression"),
  trunc = 0.01
)

Arguments

y

outcome.

d

binary treatment.

X

covariates.

regime

0/1 recommendation per unit (vector or function of X).

propensity

optional propensity vector; fitted by logit if NULL.

method

one of "aipw", "ipw", "regression".

trunc

propensity truncation.

Value

list with the policy value, SE, CI, static comparisons and diagnostics.

References

Zhang, B., Tsiatis, A. A., Davidian, M., Zhang, M. and Laber, E. (2012). Estimating optimal treatment regimes from a classification perspective. Stat, 1(1), 103-114. Murphy, S. A. (2003). Optimal dynamic treatment regimes. Journal of the Royal Statistical Society Series B, 65(2), 331-355.

Examples

set.seed(10)
n <- 400
X <- matrix(rnorm(n), n)
d <- rbinom(n, 1, 0.5)
y <- 1 + d * X[, 1] + rnorm(n)
v <- morie_regime_value(y, d, X, regime = as.integer(X[, 1] > 0))
v$value
#> [1] 1.379891