|
Activity Number:
|
260
|
|
Type:
|
Topic Contributed
|
|
Date/Time:
|
Tuesday, August 4, 2009 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
Section on Statistics in Epidemiology
|
| Abstract - #303883 |
|
Title:
|
Sufficient Cause Interactions for Categorical and Ordinal Exposures
|
|
Author(s):
|
Tyler VanderWeele*+
|
|
Companies:
|
The University of Chicago
|
|
Address:
|
Department of Health Studies, Chicago, IL, 60637,
|
|
Keywords:
|
Causal inference ; Interaction ; Sufficient Cause ; Gene-Environment ; Counterfactuals
|
|
Abstract:
|
Definitions are given for weak and strong sufficient cause interactions when the outcome is binary and the two exposures of interest are categorical or ordinal. Empirical conditions are derived for weak and strong sufficient cause interactions and these conditions are related to regression coefficients in linear and log-linear models. Weak and strong sufficient cause interactions concern cases in which some mechanism will operate under certain exposures values but will not for either some other (weak int.) or any other (strong int.) exposure values. When the exposures under consideration are binary, the notions of a weak and of a strong sufficient cause interaction coincide; however, when the exposures under consideration are ordinal or categorical, the distinction between the two concepts becomes apparent. The results are applied to two examples concerning gene-environment interactions.
|
