Marginal odds ratios: What they are, how to compute them and why sociologists might want to use them
Speaker
- Kristian Bernt Karlson, University of Copenhagen
Kristian Bernt Karlson is Associate Professor in the sociology department at the University of Copenhagen. His work revolves around educational stratification, social mobility, and quantitative methods. Recent work appears in American Journal of Sociology, Sociological Methods and Research, and Sociological Science.
Abstract
We introduce to sociologists marginal odds ratios whose main feature is that they behave like average marginal effects but retain the odds ratio interpretation. Marginal odds ratios are unaffected by rescaling or noncollapsibility bias, and they can be compared across different populations or studies. We define marginal odds ratios in terms of potential outcomes and show how to obtain an unadjusted (gross) and an adjusted (net) marginal odds ratio. We present different ways of estimating marginal odds ratios. We also present some examples using a user-written command that implements the method in Stata.