Abstract:
|
Web-based surveys and enrollment are becoming attractive approaches for epidemiologists. Participants are often self-selected, with little or no reference to a well-defined study base. This situation reactivates the discussion of the nature and importance of external validity in epidemiology. A classical approach would emphasize representativity (usually conditional on important confounders) with regard to the study base. An alternative view, led by some influential epidemiologists, claims that representativity and formal statistical transportability are irrelevant for the scientific nature of epidemiology. This view interprets 'representativity' very narrowly, as simple random sampling. The central issue is whether conditional effects in the study population, given key confounders, may be transported to relevant target populations. As emphasized by Pearl & Bareinboim, the answer will depend on the compatibility of causal structures in study and target populations and will require subject matter considerations in each concrete case. Statisticians and epidemiologists should work together to develop increased understanding of these challenges and improved tools to handle them.
|
ASA Meetings Department
732 North Washington Street, Alexandria, VA 22314
(703) 684-1221 • meetings@amstat.org
Copyright © American Statistical Association.