Matrix-Variate Discriminative Analysis, Integrative Hypothesis Testing, and Geno-Pheno A5 Analyzer

A general perspective is provided on both on hypothesis testing and discriminative analyses, by which matrix-variate discriminative analyses are proposed based on the matrix normal distribution, featured by a bi-linear extension of Fisher linear discriminant analysis and a further extension to binary variables. Moreover, a general formulation is proposed for integrative hypothesis testing and five typical categories are summarized. Furthermore, major techniques for variable selection are briefly elaborated. Finally, taking analyses of gene expression and exome sequencing as examples, we further propose a general procedure called Geno-Pheno A5 Analyzer for integrative discriminant analysis.