Question about meologit

[Cassie Wright]

Hello!

I attempting to do a mixed model ordinal logistic regression with level 2 and level 1 variables (country and individual wide variables) to determine the effect of level one and level two variables on a level two variable. I was just wondering if I should be using a random intercept for the country or individual level variables, or if there is a way to test for this.
Code:

 meologit levelonevar1 levelonevar2 levelonevar3 leveltwovar1 leveltwovar2 || country:

Thank you in advance

[Clyde Schechter]

I was just wondering if I should be using a random intercept for the country or individual level variables

I'm not sure what you are asking. Random intercepts correspond go levels of the model, not to variables. When you see a command like -me_command ... || level: variable- it is a random slope being introduced for the variable. If that is what you are asking about, the answer is best found by considering what is known about the real world data generating process. A random slope is a cross-level interaction. That is, if you specify -|| country: levelonevar2- you are saying that the effect of levelonevar2 on the outcome varies from one country to another. If this is true, and if you care about the variation among countries in this effect, than including a random slope for it at that level would make sense. Otherwise not. I should note that it is often true that the existence of a cross-level interaction is highly plausible, but that variation is not important for your research goals. In that case, it is best to omit it because random slopes are computationally expensive and can slow down your estimation.

[Cassie Wright]

Apologies for the vagueness in my question - I was confusing random intercepts with a random slope.

I am very sorry it took so long to give my thanks to you Clyde! Somehow my initial response got lost in the ether of the internet.

Best wishes, Cassie