Using Twin Study Data to Identify Causal Effects Under Latent Confounding

• Activity Number: 534 - Contributed Poster Presentations: Section on Statistics in Epidemiology
• Type: Contributed
• Date/Time: Wednesday, August 1, 2018 : 10:30 AM to 12:20 PM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Yige Li* and Zhichao Jiang and Zhi Geng
• Companies: Harvard School of Public Health and Princeton University and Peking University
• Keywords: Causal Inference; Twin Study; Gene-environment Interaction; Latent Confounding; GMM

Abstract

Twin studies, previously employed to estimate heritability in behavior genetics, are good observational studies which own some properties of randomization trials. In this paper, identifiability of average causal effects will be discussed under the presence of latent confounding and interaction of confounding, which can be interpreted as gene-environmental interactions in social science literature. Generalized method of moments estimate of average causal effects will be given and compared to the estimate from fixed effects model. An analysis of data from Chinese National Twin Registry shows that drinking amount is a significant effect modifier for the effect of cigarette smoking amount on BMI considering its interaction with genes.