Rank Conditional Coverage: Confidence Intervals for High-Dimensional Problems

Jean Morrison from the University of Chicago proposes rank conditional coverage (RCC) as a new criterion for confidence intervals in high-dimensional settings. The paper describes two bootstrap-based methods implemented in the R package 'rcc', available on CRAN, which aim to provide better coverage for the most significant parameter estimates. This work addresses the problem of marginal confidence intervals having low coverage rates for top-ranked parameters in multiple testing scenarios.

Use Cases

• Constructing confidence intervals for selected parameters in high-dimensional regression based on the proposed RCC framework.
• Evaluating coverage probabilities relative to parameter rank in multiple testing problems as described in the paper.
• Implementing bootstrap-based interval methods for high-dimensional data using the provided R package.

Strengths

• Methodology is implemented in a publicly available R package ('rcc') on CRAN.
• Addresses a specific, documented problem with marginal confidence intervals in high-dimensional settings.
• Proposes a novel coverage criterion (rank conditional coverage) backed by theoretical discussion.

Limitations

• Description metadata is limited; actual data quality requires manual inspection after download.
• Column-level documentation is absent; field semantics must be inferred after download.
• Row count and example data are unavailable, which may limit suitability assessment.

Provenance

Source

University of Chicago

Collection Method

Proposed statistical methodology and software implementation.