How can I do hypothesis test for continuous dependent variable and two categorical independent variabables which are related?

Question

[BioDH]

I am trying to analyze metagenome data.

I have two variable which I want to know whether any of them is correlated with the abundance of a specific OTU.

One is BMI (categorical) and the other is gender (categorical as well).

The dependent variable is the abundance of OTU(continuous)

Let's say the abundance of E. coli is significantly correlated with both of BMI and gender.

And I realized that BMI and gender are significantly related each other (by chi-square test).

What I am trying to know is how to know which variable is more significantly corelated with the abundance.

Is there any way to correct any effect came from the gener variable on the test between BMI and the abundance?

Any comments are welcome.

[Mensur Dlakic]

To the best of my knowledge, BMI is a number, and is therefore a continuous variable. You may choose the categories (lean, normal, overweight and obese) if that suits you better for other purposes, but it is probably easier to use it as a continuous variable when looking for this correlation.

Answer 1

[Lemire]

Use a multiple regression framework by including both sex and bmi. In R:

summary( lm( OTU ~ BMI + sex ) )

Then, in the output, the effect and significance for BMI represents an effect that is above and beyond the effect of sex (and, at the same time, the effect and significance for sex represents an effect that is above and beyond the effect of BMI). If sex is no longer significant in this framework, it means sex has no effect on OTU after accounting for BMI. If both terms are significant, then both can be viewed as independent predictors of OTU.