Hi,

I am using a slightly outdated version of QIIME2, 2018.4, but was hoping to finish the analysis in the same version of QIIME2 I began the analysis in, if at all possible.

I am analyzing data that collected multiple samples of different types from a group of infants, and I want to compare the abundance of the taxa from the different types, as while test for the influence of a few other variables such as gestational age at delivery. I am running the regression with lme-regression and setting subject ID to the grouping variable to account for repeated measures. The model runs fine, but I was a bit surprised by some of the results, namely the corrected p-value is frequently smaller than the uncorrected p-value. E.g. for the influence of antibiotic exposure on one balance, I have an uncorrected p-value of 0.889, and a corrected p-value of 0. I find this a little difficult to believe, but I am uncertain what might have caused it.

I know I included a few too many variables in my model for the sample size, but Iâ€™d planned to use a backwards elimination approach to the analysis. Unfortunately, with the corrected p-values everything appears significant. Is this an issue of an overloaded model? And does a later version of QIIME2 provided any goodness of fit testing for regression models? Or is there a chance that QIIME2-2018.4 is reporting corrected p-values that should be 1 as 0?

Iâ€™m just trying to sort out what might have gone wrong.

Thanks!