How are fixed effect p-values calculated in MaAsLin2?

Downstream analysis and statistics MaAsLin
1

I am running MaAsLin2 with model = “LM”.

For the methods section of a paper, I am trying to describe more precisely how MaAsLin2 calculates p-values for fixed effects.

I’ve been digging around the code to try to find an answer. Line 81 in fit.R runs coef(summary(fit)) where fit is made using lmerTest::lmer (line 75). Based on the lmerTest documentation, I think this means that we are calculating a p-value based on a t-statistic using “Satterthwaite’s method for denominator degrees of freedom”. Therefore, I would say something like “fixed effect p-values were calculated based on a t-statistic using Satterthwaite’s method for denominator degrees of freedom”.

Would you agree with this description?

I am digging into this because my understanding is that there is no universally recommended way to calculate p-values for fixed effects in LMMs. I have previously used the Wald test, but that is NOT what MaAsLin2 is doing, right?

Thanks in advance for your advice.

1 Like
2

Correct - MaAsLin 2 is not using a Wald test. What you’ve written looks good, and you can cite the lmerTest paper if you think it’d be helpful.

1 Like