Without any normalization and transformation, is the default LM equivalent to a parametric t-test?

Dear MaAsLin2 developer,

This might be a very silly question, I’ve been trying to understand how does MaAsLin2 work statistically, but as indicated in the title, if I don’t perform any normalization and transformation, and I only have 1 level of categorical variables to compare, the default linear model would just compare each variable with the reference using a linear model, which essentially makes it a parametric t-test, does this make sense? please correct me if I am wrong. In such a case, would MsAsLin2 tests for the assumptions of parametric tests before reporting the statistical conclusions?

Thank you!


Hi @ruizzhan,

Thanks for your question! To my understanding - if you perform no normalization or transformation (not recommended with MaAsLin) - yes these two tests would be essentially the same. Specifically, in this case, the test you are specifying, I believe, simplifies to a two-sample t-test with equal variance assumption.

I hope this helps.

1 Like

Thank you for the clarification!