11 November 2017 2 8K Report

Hi all,

Today I have read a paper published in the journal Bioinformatics on 27 Oct 2017, entitled 'An exact test for comparing a fixed quantitative property between gene sets'. This paper indicated the Mann-Whitney U (MWU) test is not compatible with known, finite populations. Thus the author proposed a novel exact test strategy for the known population median test. However, in the 'Mann-Whitney U test' subsection, the author stated that 'As such, the basic definitions, problem formulation, computations, and inferences of the MWU test are clearly nonsensical in settings in which the entire population is finite and known, such as genes in the genome'. As the samples are the whole population themselves, the randomized sampling of whole population is equivalent to the whole population itself & they should have the same statistical metrices. On the other hand , samples from known populations also could be analyzed with the MWU test. Due to my different thought about this paper, I wanna listen to your various voices.

Best,

Yanshi

Similar questions and discussions