Selecting covariates in MaAsLin2 amidst absence of statistical differences in baseline characteristics

Hello everyone,

Thank you for providing me with the fantastic tool, MaAsLin2.

I have a question regarding the selection of covariates.

I aim to perform a differential abundance test between two groups (e.g., benign vs. malignancy). Age, gender, and BMI are recognized as important factors for the development of the type of malignancy. However, the analysis of baseline characteristics between both groups shows a statistical difference only in age. In this scenario, is it appropriate to include only age as a confounding factor (fixed effect) while excluding the other known important factors?

Hi @minchance,

Unless you’re working with a very small number of samples its unlikely to impact your results to much if you include them as covariates. Without seeing the data its hard to make a strong suggestion one way or another, however, as you had indicated if there not a large difference to begin with then including them as covariates may not be needed.

Jacob Nearing

Thank you, @nearinj .

The subgroups are quite small (about 10 to 20), and whether including the pre-determined covariates or not quite impacts the results. Your comment so much helped.

Best regards,