The 'covariates' parameter in 'adjust_batch {MMUPHin}'

Thanks to you for this microbiome meta analysis tool, it’s really awesome!!
I used this tool for the first time and some details are not quite clear. Here, I have a question about The ‘covariates’ parameter in ‘adjust_batch {MMUPHin}’. In the “Performing meta-analyses of microbiome studies with MMUPHin”(Performing meta-analyses of microbiome studies with MMUPHin), the covariate option only includes “study_condition”, why not add some covariates such as age and BMI to adjust the feature abundance table? (ex: … covariates = c(“study_condition”,“age”,“BMI”) … )


Any guidance would be appreciated! Thank you!
-FQZ

This question is hard to answer in a precise way without getting into a lot of mathematical detail, but the hand-wavy answer is: we want the adjustment step to leave the variation in the outcome variable (study_condition) intact, but adjust for inconsistency in the other covariates (age, BMI) between studies. The other covariates are used to explain the outcome in the next step (when you call lm_meta()).

1 Like

Thank you so much, I know it.

I have a case where I don’t have a case-control setting, but a continuous metavariable associated with my samples, e.g. time and height. It seems that the adjustment is not working very well, if I put these as covariates. What would your recommendation be in this case? Running the adjustment without any covariates?

The covariates argument shouldn’t care if the variable is binary or continuous, but it might help if you standardized the variables first with scale().

Thanks! The scaling did not change a lot, but putting only one variable as covariate helped a lot.