Estimate restricted mean time in favor of treatment from ordered state paths. Lower numeric states denote better outcomes. Estimation is available for complete paths and paths subject to independent censoring; inference includes projection standard errors, normal intervals, two-sided tests, and stratified bootstrap methods that re-estimate the censoring distribution.
Install the source archive with base R:
install.packages("RMTIF_0.3.1.tar.gz", repos = NULL, type = "source")
library(RMTIF)For example, obtain projection inference from a matrix of pairwise net times whose rows identify control subjects and columns identify treatment subjects:
net <- matrix(c(1, 2, 3, 4), nrow = 2, byrow = TRUE)
rmtif_projection_inference(net, conf_level = 0.95)The R source is copied byte for byte from the accompanying
functions_v1.2.R. The separate examples_v1.2.R
contains manuscript design settings and writes the manuscript tables.
Long manuscript Monte Carlo runs are excluded from package examples and
tests. Function help includes concise examples using independently
chosen illustration inputs.
The restricted-time estimand follows Mao (2023), “On restricted mean time in favor of treatment”, Biometrics 79, 61-72, doi:10.1111/biom.13570. The pairwise regression is motivated by Mao and Wang (2021), “A class of proportional win-fractions regression models for composite outcomes”, Biometrics 77, 1265-1275, doi:10.1111/biom.13382.
The manuscript describing this package’s reversible-path extensions is unpublished and currently has no public URL or DOI. The references above identify the published foundations; package help documents the implemented estimators and functions.
Study functions return summary, replicate, design, and runtime data frames. They run quietly in both serial and parallel execution. Users can explicitly print or save the returned information they need.