“Hypothesis tests of equivalence are typically known for their application in bioequivalence studies and acceptance sampling. Their application to gene expression data, in particular high-dimensional gene expression data, has only recently been studied,” observed Dr. Xiangqin Cui, associate professor in the department of biostatistics, section on statistical genetics, and Dr. Celeste Yang, former doctoral student and graduate research assistant in the department of biostatistics, at the University of Alabama at Birmingham, in a recent paper that studied how the F-test and range test function pertaining to microarray expression data. Co-investigator in the study was department colleague Dr. Alfred A. Bartolucci, professor emeritus.
The multigroup equivalence tests were adapted to the difference ratio, which is an accepted equivalence criterion. Study results indicate that although both the F-test and range test can attain moderate power — “controlling the type I error at nominal level for typical expression microarray studies with the benefit of easy-to-interpret equivalence limits” — the F-test proved to be the more powerful of the two. However, powers were found to be similar when three groups were compared. The researchers noted that “the two multigroup tests were applied to a prostate cancer microarray dataset to identify genes whose expression follows a prespecified trajectory across five prostate cancer stages.”
“Multigroup Equivalence Analysis for High-Dimensional Expression Data” was published in November in the journal Cancer Informatics.