Fixed vs random effects

Downstream analysis and statistics MaAsLin
1

Hi, I am trying to account for random drift in my microbiota samples.

So I have 2 groups: a treatment group with 2 timepoints and a control group also with 2 timepoints. I am trying to show that taxa that are significantly changed between 2 timepoints in my treatment group is not due to random drift.

In MaAsLin2, can I place time as fixed effects, and treatment as random effects? Maybe to include subject as random effects too?

2

Hi -

I’d say use time and treatment as fixed effects. Time because you are interested in testing for time change. Treatment because you’d usually expect for it to have an effect. Then you can use subject as random effect (i.e. grouping variable) - observations across time points for the same subject should be correlated, and this accounts for that.

Best,
Siyuan

1 Like
3

ok thanks a lot Siyuan!