0
$\begingroup$

I am using R Studio to run logistic regressions. I have a dataset, which analyses voting behaviour. I want to analyse the effect of different independent variables on the dependent variable "participation in election". Moreover I want to see if there are differences between the subgroup of the first time voters and the group in general.

Since the coefficients of logistic regressions are difficult to interpret, I wanted to interpret the Average marginal effects of the models. I read about different ways of doing so but I cant find a solution that fits my plan completely. Can I compare the AMEs of the two separate logistic regression models, one for the whole group and one for the first time voters and interpret the differences between the AMEs of the two models or is that statistically wrong? Or is there another way for my problem?

Improve this question
$\endgroup$
2
  • $\begingroup$ Marginal effects are much more difficult to interpret and depend on the both the covariate distributions in the sample and in the population. What's wrong with the simple intepretable compare like-with-like effects that come automatically out of logistic regression? fharrell.com/post/robcov $\endgroup$ Commented Aug 21, 2022 at 11:19
  • $\begingroup$ I feel like logit coefficients aren't always directly comparable across models (doi.org/10.1016/j.ssresearch.2022.102802), although not sure if that particular issue is applicable here. I also think that many people find logit coefficients and odds ratios very challenging to interpret because we don't have a good intuition about odds (or especially log odds). We're much more used to thinking in terms of changes in probabilities, which is something that (despite their problems) marginal effects allows. $\endgroup$ Commented May 7 at 19:16

1 Answer 1

Reset to default
0
$\begingroup$

It's not clear from your question why you would need to run two separate models. Maybe if you provide an example of your data it will be clearer?

I would include a binary variable indicating whether or not a voter is a first-time voter or not as a predictor in a single model, then simply look at the model summary to see if there's a significant effect of that predictor.

Improve this answer
$\endgroup$

Start asking to get answers

Find the answer to your question by asking.

Ask question

Explore related questions

See similar questions with these tags.