|
Abstract:
|
The central goal of many studies is to estimate the effect of some "treatment." Randomized experiments provide unbiased estimates of treatment effects, but are not always feasible. When randomization is not possible, observational studies are used instead. As Gertrude Cox stressed the need for good design of experiments, observational studies also need to be carefully designed. This includes thinking very carefully about the questions of interest, the data sources, and the comparisons made. This talk will focus on the design of observational studies, with examples from mental health, education, and health care. Experimental and observational study designs are increasingly important in these fields given recent movements for accountability and "evidence-based practices," and provide a way for statistical methods to influence policy decisions.
|
