Marginal Odds Ratios: What They Are, How to Compute Them, and Why Sociologists Might Want to Use Them
Open Access
- 1 January 2023
- journal article
- research article
- Published by Society for Sociological Science in Sociological Science
- Vol. 10, 332-347
Abstract
As sociologists are increasingly turning away from using odds ratios, reporting average marginal effects is becoming more popular. We aim to restore the use of odds ratios in sociological research by introducing marginal odds ratios. Unlike conventional odds ratios, marginal odds ratios are not affected by omitted covariates in arbitrary ways. Marginal odds ratios thus behave like average marginal effects but retain the relative effect interpretation of the odds ratio. We argue that marginal odds ratios are well suited for much sociological inquiry and should be reported as a complement to the reporting of average marginal effects. We define marginal odds ratios in terms of potential outcomes, show their close relationship to average marginal effects, and discuss their potential advantages over conventional odds ratios. We also briefly discuss how to estimate marginal odds ratios and present examples comparing marginal odds ratios with conventional odds ratios and average marginal effects.Keywords
This publication has 32 references indexed in Scilit:
- Total, Direct, and Indirect Effects in Logit and Probit ModelsSociological Methods & Research, 2013
- Comparing Regression Coefficients Between Same-sample Nested Models Using Logit and ProbitSociological Methodology, 2012
- Estimators and confidence intervals for the marginal odds ratio using logistic regression and propensity score stratificationStatistics in Medicine, 2010
- Logistic Regression: Why We Cannot Do What We Think We Can Do, and What We Can Do About ItEuropean Sociological Review, 2009
- Unconditional Quantile RegressionsEconometrica, 2009
- Estimating a Marginal Causal Odds Ratio Subject to ConfoundingCommunications in Statistics - Theory and Methods, 2008
- Robustness of Logit Analysis: Unobserved Heterogeneity and Mis‐specified Disturbances*Oxford Bulletin of Economics and Statistics, 2007
- Comparing Logit and Probit Coefficients Across GroupsSociological Methods & Research, 1999
- A new approach to causal inference in mortality studies with a sustained exposure period—application to control of the healthy worker survivor effectMathematical Modelling, 1986
- Change and Stability in Educational StratificationAmerican Sociological Review, 1981