search

ggloor / ALDEx_bioc

ALDEx_bioc is the working directory for updating bioconductor
27 stars 13 forks source link

Effect_value_interpretation #39

Closed StefanoCast closed 3 years ago

StefanoCast commented 3 years ago

Hi, I am just starting to using ALDEx2 on my metagenomic datasets: it is really easy to implement! I am wondering about the interpretation of "effect" values, as calculated by the "aldex" function in a two group comparison of a metagenomic feature table.
For example, a feature with "rab.all = -2.7", "rab.win.0 = -1.21", "rab.win.1 = -9.3", "diff.btw= -7.9", "diff.win = 3.4" and "effect = -2.1", is inferred to be highly abundant in the class "0" (given that "effect" is negative), or not? Corrected p-values for the feature are both < 0.05. Sample size of class "1" is 21, while size of class "0" is 13.

Then, does a positive effect mean a major abundance of such feature in the class "1"? Which is the optimal "effect" cutoff (in absolute value)? Should I at the same time consider the "overlap" and corrected p-values for the identification of differentially abundant taxa? Thanks a lot, best regards.

Stefano

ggloor commented 3 years ago

Hi Stefano I'm attaching a preprint that will be submitted soon that will answer this question.

TLDR: an absolute effect of >2 is highly different and will almost certainly reproduce if you replicate the experiment; i.e., it is a real difference, and a very large one!

peerj-gloor.pdf

StefanoCast commented 3 years ago

Thank you very much for the reply, your manuscript is very helpful!

Stefano