Adjusting covariate in regression model

Question

[Jonathan Yoou]

Hi all,

I'm analysing a microbial abundance data with covariates of gender and living places. What I want to know is to know the genus significantly changing following different living place regardless of host sex, and the model in my mind is like: Abundance~living_place + gender.

Here is my question: I have several ways to do this in my mind like below:

1. Run the model and choose genus showing significance only on living_place
2. Stratification on gender (i.e divide the dataset into two, one for male and the other for female, and build 2 models for each)
3. Making a mixed effect model to randomize gender effect (i.e abundance ~ living_place + (1|gender)

I thought option number 2 is the most conservative and proper way to answer my question, but I wonder if I can handle with building just one model and simplify the procedures. If it is not a gender, but multi-categorical variable, it becomes complicated, you know.

Thank you for your interest in advance!

[Chris Dean]

(1) is fine, but you should probably check for an interaction first; (2) should do this if there is an interaction; (3) including gender as random effect does not seem appropriate

[Jonathan Yoou]

Thanks a lot with your comment!