Software
Our methods are released as documented, open-source packages — on CRAN for R, and as installable commands for Stata — so the procedures we publish can be applied directly to real trials and observational studies.
R packages
v1.2.0 · 2025-12-04 · Uno H, Horiguchi M, Qian Z
Survival data analysis using average hazard. Performs two-sample comparisons based on average hazard with survival weight (AHSW) or the censoring-free incidence rate (CFIR) proposed by Uno and Horiguchi (2023).
Tools for the design and analysis of randomized clinical trials with average hazard, accompanying Horiguchi et al. (2026).
Average hazard summaries computed from adjusted survival curves — adjusted average hazard with survival weight (AHSW), restricted mean survival time, and event rate. It takes an adjusted survival curve object, and supplies an AIPTW estimator of its own for that step.
pak::pak("kkevin821/survAHadjust")
v1.0-4 · 2022-06-14
Two-sample comparisons using the restricted mean survival time (RMST) as a summary measure of the survival time distribution. Computes three contrast metrics — difference in RMST, ratio of RMST, and ratio of restricted mean time lost — with ANCOVA-type covariate adjustment as well as unadjusted analyses.
v1.1.0 · 2023-03-28
Implements the procedure of Horiguchi et al. (2018): specify a set of truncation time points τ for calculating RMST, test for equality, and estimate the between-group difference at those τ, accounting for the multiplicity introduced by specifying several τ.
v0.1.0 · 2020-06-04
Permutation tests for the difference in RMST between groups — useful when the sample size is under about 50 per group, where the usual asymptotic test can show non-negligible type I error inflation. Implements the six solutions examined by Horiguchi and Uno (2020) for permutations in which a survival function cannot be defined, and reports the asymptotic result for reference.
v0.1.1 · 2017-05-18
Power and sample size calculations based on the difference in restricted mean survival time.
v1.0-5 · 2017-06-23
Inference for several model-free group contrast measures, including differences and ratios of cumulative incidence rates at given time points, quantiles, and RMST. Includes two covariate adjustment procedures (regression and augmentation) for the RMST-based metrics.
v1.0.1 · 2022-06-11
Tests for equality of two survival functions based on integrated weighted differences of two Kaplan–Meier curves.
v1.0-3 · 2021-02-10
Inference for the C-statistic of risk prediction models with censored survival data, using the method of Uno et al. (2011). Inference for the difference in C between two competing models is also implemented.
v1.1-2 · 2022-04-23
Inference for a class of measures comparing competing risk prediction models with censored survival data, including the integrated discrimination improvement index (IDI) and the category-less net reclassification index (NRI).
Graphical procedures for evaluating overall and subject-specific incremental values from new predictors with censored event time data.
Evaluating prediction rules for t-year survivors.
Stata commands
Install directly from within Stata — no manual download required. Both packages are MIT-licensed and archived on Zenodo.
Xing E, Tian L, Uno H · MIT
Survival analysis using average hazard with survival weight (AH) — a robust, interpretable alternative to hazard-ratio-based comparisons. Supports one-sample estimation and two-sample comparisons, both unstratified and stratified; stratified analyses use direct standardization with stratum-size weights by default, and user-supplied weights are also accepted.
net from "https://uno1lab.com/stata-stah/" net install stah
Cronin A, Tian L, Uno H · MIT
Restricted mean survival time (RMST) analysis in Stata. Computes RMST
differences and ratios and supports ANCOVA-type covariate adjustment for
k-sample comparisons; the companion command strmst2pw
summarizes pairwise comparisons for the adjusted analyses.
net from "https://uno1lab.com/stata-strmst2/" net install strmst2
Distributed on GitHub
Released code that sits outside the groups above, distributed through github.com/uno1lab.
Decomposing the variance of cluster-specific proportions for binary outcome data, as used in our cancer care disparity work.
SAS implementation of the RECUR algorithm, which determines the timing of cancer recurrence from claims or electronic medical record data.
