IHW

IHW assigns data-driven weights to hypothesis tests using covariates to increase power in large-scale multiple testing while controlling the false discovery rate.

Key Features:

• Data-driven weighting: Assigns weights to individual hypothesis tests based on covariates.
• Covariate independence: Uses covariates that are independent of p-values under the null hypothesis to preserve validity.
• Input format: Processes a two-column input table consisting of p-values and associated covariates.
• Covariate types: Supports continuous-valued and categorical covariates.
• Power and prior information: Leverages covariates informative about the power of each test or the prior probability that the null hypothesis is true.
• False discovery rate control: Controls the false discovery rate while enhancing detection sensitivity.
• Comparison to Benjamini-Hochberg: Increases power relative to the Benjamini-Hochberg procedure in large-scale testing settings.
• Prioritization of hypotheses: Prioritizes hypotheses that are more likely to yield significant results based on covariates.

Scientific Applications:

• Genomics: Applied to large-scale hypothesis testing problems in genomics.
• High-throughput biology: Used in high-throughput biology experiments involving many simultaneous tests.
• Large-scale multiple testing: Suited to extensive datasets where discovering true associations among many tests is challenging.

Methodology:

Processes a two-column table of p-values and covariates (continuous or categorical), assigns data-driven weights to each hypothesis based on covariates that are independent of p-values under the null, and controls the false discovery rate to increase testing power relative to the Benjamini-Hochberg procedure.