Covariance Adjustment in Randomized Experiments and Observational Studies

摘要: By slightly reframing the concept of covariance adjustment in randomized experiments, a method exact permutation inference is derived that entirely free distributional assumptions and uses random assignment treatments as "reasoned basis for inference." This may be used with many forms adjustment, including robust regression locally weighted smoothers. The then generalized to observational studies where were not randomly assigned, so sensitivity hidden biases must examined. Adjustments using an instrumental variable are also discussed. methods illustrated data from two studies.