GitLab

comat now deals with NA (NA or Inf forbidden in intial matrices but some NA are ok afterward, notably in heatmaps)

comat can now deal with negative matrix distances for p value computation (interest in knowing if the tendency is toward matrix 1 or matrix 2

new comat operational (possibility to fix margins and sampling before normalization for p value and MDS CI computation), remains to correct the sampling for MDS CI

new function normalize.fun() created, remain to add this function after sampling the raw integer matrices

solved: 1) now I can reproduce the Nat Biotech results again. See C:\Users\Gael\Documents\Hub projects/20170704 Pierre Bruhns 8528/labbook_gael_PB_8528.docx section '20200517 Big problem with Comat'; 2) TT1 can be normalized several times (depending on the comparison couples) -> now saved normalized matrices are not overwritten)

going to v4 by adding fixed margins for the p values computations / comat_subsampling_by_cell_nb_not_good being a way to have fast sampling but that does not lead to the results with the older method / current code achieved but I have to confirm that we obtain the same results as with v2