LBE

LBE estimates the false discovery rate (FDR) in microarray gene expression studies by introducing a family of estimators for pi0 to reduce bias and improve FDR accuracy.

Key Features:

• Marginal p-value framework: Operates within procedures that rely on the marginal distribution of p-values without assuming a specific form for alternative hypotheses.
• Novel pi0 estimators: Introduces a family of estimators for pi0, the probability that a gene is not differentially expressed.
• Bias reduction: Addresses the conservative bias of traditional pi0 estimators to facilitate more accurate FDR calculations.
• Finite-sample performance: Simulation studies report low mean square error in finite samples relative to QVALUE, BUM, and SPLOSH.
• Applied validation: Has been applied to real microarray datasets to assess statistical inference performance.
• R implementation: Provided as an R function for integration into analysis workflows.

Scientific Applications:

• Microarray gene expression analysis: Estimating FDR and controlling false positives in microarray experiments.
• High-throughput experiments: Applicable to high-throughput biological studies that require control of the expected proportion of false discoveries.
• Method comparison and benchmarking: Used to compare pi0 and FDR estimation methods via simulation studies against QVALUE, BUM, and SPLOSH.

Methodology:

Uses the marginal distribution of p-values without assuming specific alternative-hypothesis forms, applies a novel family of pi0 estimators, and is evaluated via simulation studies comparing mean square error to QVALUE, BUM, and SPLOSH; implemented as an R function.