Estimation of a Marginal Causal Odds Ratio in a Matched Case-Control Design

• Activity Number: 139
• Type: Contributed
• Date/Time: Monday, August 1, 2011 : 8:30 AM to 10:20 AM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Emma Persson*+ and Ingeborg Waernbaum
• Companies: Umeå University and Umeå University
• Keywords: causal inference ; retrospective sampling ; confounding

Abstract

A common practice in the analysis of the effect of an exposure of interest on the development of a rare disease using case-control data is the estimation of odds ratios. A conditional odds ratio can help a clinician to decide whether or not a treatment is beneficial for a particular patient, while a marginal odds ratio can be used to assess the effect of a treatment in the population as a whole. Whereas statistical development has to a large extent focused on the former the latter is a parameter more relevant for decision makers. In this paper we compare estimators of the marginal causal odds ratio. Recently, for matched case control designs, targeted maximum likelihood estimators of marginal causal parameters have been proposed. Here, comparisons are made to standard estimators of the odds ratio. Also, an estimator of the marginal causal odds ratio for unmatched case-control designs is proposed. The estimator is based on an intercept adjusted logistic regression model. The finite sample performances of the estimators are highlighted in simulations. The estimators are also applied to data where the effect of socioeconomic variables on the risk of type 1 diabetes is studied.