0

I used spearman correlations to test associations between relative abundances (percentages of each microbe present in each sample) of certain microbes and continuous variables. Many people have zero percent of certain microbes which is normal for the microbiome. What type of regression can I use to follow up on the spearman correlations that can use zero-inflated percentage data? I initially used Negative Binomial regressions but that forced me to switch to absolute abundances (count data) as opposed to relative abundances (percentages).

kjetil b halvorsen
  • 77,844
Desiree Chase
  • 1
  • You can use offsets with negative binomial regression to keep the interpretation as rates (percentage) – kjetil b halvorsen Jan 13 '23 at 15:28
  • This seems to be very similar to https://stats.stackexchange.com/questions/31300/dealing-with-0-1-values-in-a-beta-regression. Although you haven't settled on beta regression here, it seems likely to me that it will be appropriate for your problem. – jbowman Jan 13 '23 at 17:32
  • Thank you for your reply, kjetil-b-halvorsen, do you happen to have an R code that I can use to do this? Do I first offset the percentages and then plug them into the negative binomial regression? – Desiree Chase Jan 16 '23 at 23:57
  • jbowman, thanks for your input. My data are zero-inflated percentages. Would I need to transform the data first, prior to calculating the beta regression? If so, which transformation? – Desiree Chase Jan 17 '23 at 00:00

0 Answers0