Skip to contents

Analyze Power for RMST Model with Covariate-Dependent Censoring (Analytic)

Source: R/dependent_censoring_analytical.R
DC.power.analytical.Rd

Performs power analysis for an RMST model when the censoring mechanism depends on observed covariates. Competing risks are not modeled here.

Usage

DC.power.analytical(
  pilot_data,
  time_var,
  status_var,
  arm_var,
  sample_sizes,
  linear_terms = NULL,
  L,
  alpha = 0.05
)

Arguments

pilot_data

A data.frame with pilot data.

time_var

Name of the time-to-event variable.

status_var

Name of the primary event indicator (1=event, 0=censored).

arm_var

Name of the treatment indicator (1=treatment, 0=control).

sample_sizes

Numeric vector of per-arm sample sizes for power.

linear_terms

Optional character vector of additional covariate names (used in both models).

L

RMST truncation time.

alpha

Two-sided Type I error.

Value

A list with:

results_data

A data.frame with N_per_Arm and Power.

results_plot

A ggplot object of the power curve.

Details

This function assumes a single censoring process whose hazard depends on covariates (but not on treatment by default). It fits a Cox model for censoring $$\Pr(\text{censoring by } t \mid X) = 1 - G(t \mid X)$$ using Surv(time, status==0) ~ linear\_terms, then forms inverse-probability-of-censoring weights (IPCW) \(w_i = 1/\hat G(Y_i\mid X_i)\) evaluated at \(Y_i=\min(T_i,L)\). The RMST regression \(E[Y_i \mid A_i,X_i]\) is then fit by weighted least squares, and power is derived from a sandwich variance that ignores uncertainty from estimating \(\hat G\).

Note: dep_cens_status_var is accepted for API compatibility but is ignored under this setting (no competing risks are modeled).