Randomization-Based Nonparanormal Adjusted Marginal Inference

Adjusting for prognostic baseline covariates in the analysis of randomized controlled trials is recommended to increase the precision of treatment effect estimates. Such adjustment, however, introduces a complication known as noncollapsibility: for effect measures such as the log-odds or log-hazard ratio, the estimated treatment effect becomes conditional instead of marginal. Methods that target a marginal estimand even after covariate adjustment have been developed, but typically rely on population-based inference, which assumes patients are randomly sampled from some larger population, rarely true in practice. Such methods have been shown to inflate type I error, particularly in small samples or when many covariates are adjusted for. Inference based on randomization requires no such assumption, as it relies on the randomization scheme, which is known exactly, yielding a test that is exact even in finite samples. We present a randomization-based test for covariate-adjusted marginal treatment effects, built on nonparanormal adjusted marginal inference, applicable to continuous, binary, ordinal, and time-to-event outcomes. In simulation studies, this test controls type I error where the standard Wald test does not. Applied to 20 historical phase III time-to-event randomized trials, the covariate-adjusted estimator achieves precision gains comparable to an existing competitor method.

Authors

Publication Details

Journal
arXiv (Cornell University)
Published
2026-11-17
DOI
https://doi.org/10.48550/arxiv.2611.22542
Primary Topic
Statistical Methods in Clinical Trials
Type
preprint
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
preprint

Randomization-Based Nonparanormal Adjusted Marginal Inference

Susanne Dandl, Mark Baillie, Leticia Wüthrich
arXiv (Cornell University)
Statistical Methods in Clinical Trials
preprint

Randomization-Based Nonparanormal Adjusted Marginal Inference

Susanne Dandl, Mark Baillie, Leticia Wüthrich
preprint en

Abstract

Adjusting for prognostic baseline covariates in the analysis of randomized controlled trials is recommended to increase the precision of treatment effect estimates. Such adjustment, however, introduces a complication known as noncollapsibility: for effect measures such as the log-odds or log-hazard ratio, the estimated treatment effect becomes conditional instead of marginal. Methods that target a marginal estimand even after covariate adjustment have been developed, but typically rely on population-based inference, which assumes patients are randomly sampled from some larger population, rarely true in practice. Such methods have been shown to inflate type I error, particularly in small samples or when many covariates are adjusted for. Inference based on randomization requires no such assumption, as it relies on the randomization scheme, which is known exactly, yielding a test that is exact even in finite samples. We present a randomization-based test for covariate-adjusted marginal treatment effects, built on nonparanormal adjusted marginal inference, applicable to continuous, binary, ordinal, and time-to-event outcomes. In simulation studies, this test controls type I error where the standard Wald test does not. Applied to 20 historical phase III time-to-event randomized trials, the covariate-adjusted estimator achieves precision gains comparable to an existing competitor method.

arXiv (Cornell University)
Statistical Methods in Clinical Trials
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon