Relation of W statistic and beta coefficient about ancombc HOT 2 CLOSED

Hi Rudy,

That is a great question!

You are absolutely right, the log-linear model you found in the paper is equivalent to the model you have written down, which evolves other covariates x. In our original paper, our focus was group comparisons. However, we were aware of the need for covariate adjustment, as well as pairwise group comparison, repeated measures, etc. So we believe the general log-linear model is more convenient and we will use these notations in the upcoming ANCOM-BC extension paper. The R user manual just took a step ahead :)

As shown in the user manual, W is the standardized log fold-change (beta), which is a Wald-type of statistic and the p-value is computed by assuming Nomarlity.

Best,
Huang

from ancombc.

Dear Huang,

Thank you for your response, it is somewhat more clear to me now. Looking forward to the upcoming ANCOM-BC extension paper!

Good luck,
Rudy

from ancombc.