test for no change in population after policy (with categorical variables)

[Fabio Martinenghi]

Hi all,

I am trying to test whether the class of people in a marriage are the same after a policy. The share of people in a marriage, I found, is unchanged by the policy. Now I want to know if they belong to the same socio-demographics.

I am working mostly with discrete/categorical variables. Do you have any advice on how to proceed? I am looking for an equivalent of a paired samples T-test, but for categorical variables. I know a Chi-squared test and the McNemar's test can be used for binary variables; a t-test or a logistic regression for continuous ones. I read you can extend McNemar's test to multilevel categorical variables and possibly run a multinomial logit, but I would like further advice on this.

Thanks!

[Maarten Buis]

So you are talking about a three way table: policy by marriage by class That sounds like a log-linear model: https://www.stata.com/meeting/uk15/a..._loglinear.pdf

[Fabio Martinenghi]

Yes, it is exactly what I was looking for! I am very happy I decided to write this post. I basically have to build this table once for every class I want to test. Thank you so much Maarten.