DAtest Versions Save

• Include ANCOM-BC

Include t-test with rank-normalization - Described here: https://doi.org/10.1093/bib/bbab059

• 2.7.17: Make safer check of input class
• 2.7.16: Fix bug with parallel in R 4.0 in MacOS RStudio

No longer export internal functions Change name of internal function

Changes in this version includes the following:

• Examples added to most functions
• Value sections added
• set.seed removed internally, such that it has to be set externally for reproducibility
• Minor bug fixes

Main changes in this version were made to pass R CMD Check

Other changes:

• RAIDA not included anymore, due to instability
• mvabund not included by default anymore due to slowness

• Added 8 new methods: ALR and CLR normalization for t-test, ANOVA, linear models, and limma
• Fixed compatibility problems with baySeq
• Removed ancom and the groupSig function
• Included checks for random-effects and zero-inflated models.
• Renamed Spike.detect.rate to Power
• mvabund is included by default and 20 runs is set as default in testDA

Fix bugs related to new versions of phyloseq and snow packages

Fix error with normalization in DA.per, DA.llm2 and DA.ltt

Changed the Score so that methods with high FDR come out worse: Score = (AUC-0.5) * Spike.detect.rate - FDR

testDA now also outputs False Discovery Rate plot.DA plots FDR, Spike.detect.rate and AUC (not FPR anymore) FPR is set to NA for ANCOM and SAMseq as these do not output nominal p-values

A Score with confidence limits is now computed in the summary function to compare the methods. The score is AUC * Spike.detect.rate - FDR and can range between -1 and 1. Methods are ranked by this Score in both the plot and summary function.