Inverse probability weighted (IPW) estimator for regression parameters

Fits a regression model using inverse probability weighting, where the weights are derived from a parametric model for the censoring distribution C | (Y, Z, ...). The user supplies a right-hand-side formula (e.g. ~ y + Z) which is used to model Surv(W, 1 - D) via a Weibull AFT model. The resulting estimated survival probabilities at W are used as weights in the IPW estimating equations.

Usage

estimate_beta_ipw(data_yXZ, model, model_weights)

Arguments

data_yXZ

A data frame containing the variables in model, as well as W (observed min(X, C)), D (indicator I(X <= C)), and the covariates appearing in model_weights.

model

A formula specifying the outcome regression model (e.g. y ~ AW + Z).

model_weights

A right-hand-side formula specifying the variables in the censoring model, e.g. ~ y + Z. This will be expanded to Surv(W, 1 - D) ~ y + Z internally.

Value

A list with components

beta_est

A 1 x (p + 1) matrix of regression coefficients followed by the residual standard deviation (last element).

se_est

A 1 x p matrix of estimated standard errors for the regression coefficients.