Can I use a logistic regression to test the presence/absence of phylum between healthy-diseased?

Question

[pple2202]

Hi, all.

I'm analyzing the OTU table now and want to know whether the phylum specifically exist in the Healthy or the Diseased.

To do so, I carried out a Logistic Regression with binary dependent variable (1: Presence, 0: Absence) and State (Healthy - Lightly Diseased, Healthy - Heavily Diseased) as explanatory variable.

Do you think it makes sense? If not, what else I can use??

Thank you all :)

[Jean-Karim Heriche]

Why not do an association test?

[pple2202]

Hey, thank you for your reply. I thought regression would be more practical and commonly used so that's why I used it. Would it be better with using Chi-square test for association test?

[Jean-Karim Heriche]

That would be my first choice and is actually simpler than doing regression.

[pple2202]

Thanks a lot for your comment :)

[pple2202]

Hi, can I ask one more? I get the statistical significance of their presence but how can I know their direction? Like which one is more likely to be present in the Healthy or diseased?

[Jean-Karim Heriche]

You can quantify the association from the values in the contingency table using the odds ratio or the relative risk.