Schedule and Presentations 2016
Finding preventive interventions or treatments for cognitive decline and dementia requires that we identify important predictors of cognitive outcomes, and also evaluate whether those predictors are causal. Research on tools to strengthen causal inferences has flourished in recent decades, but those tools are not yet standard in research on cognitive aging. The 2016 Friday Harbor conference will focus on clearly articulating the assumptions necessary for drawing causal inferences from observational evidence and applying appropriate analyses to draw causal inferences under those assumptions. We emphasize four types of tools or analytic approaches that hold promise in our field: (1) instrumental variables (IV) analyses, including policy-based natural experiments and genetic “Mendelian Randomization” studies; (2) propensity score based methods, including inverse probability weighting to estimate marginal structural models for complex treatment regimes and selective attrition or survival; (3) mediation models using modern tools to decompose total effects into direct and indirect effects; and (4) simulation-based bias analyses.
A major part of the program will be devoted to data analyses that address specific scientific hypotheses relevant to this year’s theme. Workgroups will apply causal inference methods to address important questions about cognitive aging. Didactic presentations will introduce the causal inference approaches and methods that will be used in the workgroups. Workgroups continue to interact after the conference to follow these analyses to completion of scientific manuscripts.