POWER-ENHANCED MULTIPLE DECISION FUNCTIONS CONTROLLING FAMILY-WISE ERROR AND FALSE DISCOVERY RATES

Area of Science:

• Statistical methodology
• Biostatistics
• Data science

Background:

• Multiple hypotheses testing is crucial in analyzing complex datasets.
• Existing methods like Šidák and Benjamini-Hochberg (BH) control error rates but may not be optimal.
• High-throughput technologies generate large datasets requiring efficient statistical analysis.

Purpose of the Study:

• To develop improved procedures for multiple hypotheses testing.
• To achieve better control of family-wise error rate (FWER) and false discovery rate (FDR).
• To reduce the missed discovery rate (MDR).

Main Methods:

• Developed novel procedures exploiting differences in individual test powers.
• Utilized a decision-theoretic framework.
• Incorporated auxiliary randomizers for discrete or mixed-type data and nonparametric tests.

Main Results:

• Achieved smaller missed discovery rates (MDR) compared to existing procedures.
• Demonstrated the advantage of incorporating individual test powers into decision functions.
• Showcased the utility of receiver operating characteristic (ROC) functions in enhancing testing procedures.

Conclusions:

• Proposed procedures offer superior performance in multiple hypotheses testing.
• Considering individual test powers is essential for more effective error rate control.
• The methods are applicable to high-dimensional data analysis across various scientific fields.