Statistical Methodology
Two-Phase Sampling Designs Extended to Ordinal Outcomes Enable Cost-Efficient Biomarker Studies
Two-phase sampling designs, which reduce costs by restricting expensive measurements (e.g., biomarkers, genomic data) to an informative subset of participants, have been extended for the first time to studies with ordinal outcomes. The authors propose three outcome-informed phase 2 selection strateg…
Anytime-Valid Sequential Tests Enable Real-Time Monitoring of Stochastic Dominance
The paper introduces nonparametric e-processes for sequential, anytime-valid tests of stochastic dominance (SD), distinguishing from mean dominance by comparing full distributions. For first-order SD, these e-processes mix asymptotically growth-rate optimal e-variables, achieving power one. The fram…
Exact Study-Based Decomposition of Direct and Indirect Evidence in Network Meta-Analysis via Contrast-Space Projection
The method reformulates network meta-analysis (NMA) in contrast space as an orthogonal projection of observed pairwise treatment contrasts onto the consistency-constrained subspace, yielding an explicit linear mapping that reproduces NMA estimates. It defines direct and indirect evidence through a c…
Bayesian Hybrid Shrinkage Mitigates Winner's Curse in Online Experimentation
The "Winner's Curse" in online experimentation leads to inflated treatment effect estimates due to using the same data for selection and evaluation. This bias is influenced by sampling variability, selection thresholds, and true effect size. A novel Bayesian approach, incorporating local shrinkage f…
Two GPC-Based Methods Reliably Control Type I Error in Stepped-Wedge Cluster Trials
Stepped-wedge cluster randomised trials (SW-CRTs) pose analytical challenges when composite endpoints are evaluated using generalized pairwise comparisons (GPC), as most estimators fail to adequately account for clustering and temporal trends. A comprehensive simulation study across varying ICCs, cl…
Time Series Gaussian Chain Graph Models for Blockwise Dependencies
This paper introduces a novel class of time series Gaussian chain graph models designed to characterize both contemporaneous and lagged causal relationships within and across partitioned blocks in multivariate time series data. The approach utilizes a cross-frequency shared group sparse plus group l…