Nonparametric relevance-shifted multiple testing procedures for the analysis of high-dimensional multivariate data...

Area of Science:

• Biostatistics
• Genomics
• Bioinformatics

Background:

• Many research areas require detecting differences between treatment groups across multiple variables.
• Standard tests often focus on deviations from zero or one, which may lack biological relevance.
• Biologically meaningful differences are crucial in fields like microarray data analysis.

Purpose of the Study:

• To address the need for relevance-shifted tests on ratios in multivariate two-sample group designs.
• To develop empirical procedures for identifying biologically relevant differences.
• To provide methods that account for the entire parameter space between relevance limits.

Main Methods:

• Proposes two empirical procedures embedding relevance-shifted tests on ratios.
• Incorporates multiplicity correction to address the multiple testing problem inherent in analyzing each variable.
• Extends existing procedures for point null hypotheses to control the familywise error rate exactly.

Main Results:

• Presents two novel procedures for relevance-shifted tests on ratios.
• Ensures exact control of the familywise error rate.
• Handles the complexity arising from shifted null hypotheses in both directions.

Conclusions:

• The first procedure uses a permutation algorithm suitable for moderately large sample sizes.
• The second procedure is advantageous for experiments with limited sample sizes.
• The second procedure employs data-driven hypothesis ordering for multiplicity correction.