Sensitivity Analysis for Observational Data: Method and Computation
*Bo Lu, Ohio State University
Keywords: Observational Studies, Sensitivity Analysis, Matching
Unlike randomized experiments, treated and control groups may not be comparable at baseline in observational studies. Baseline differences that have been accurately measured in observed covariates can often be removed by matching, stratification or model based adjustments. However, there is usually the concern that some important baseline differences were not measured, so that individuals who appear comparable may not actually be. A sensitivity analysis in an observational study addresses the question what the unmeasured covariate would have to be like to alter the conclusions of the study. This course will start with Cochran’s famous example of the association between smoking and lung cancer, and introduce the methodology formulated by Rosenbaum for the sensitivity analysis with matched or stratified data. The course will also cover the implementation of sensitivity analysis using statistical software package R and illustrate the method with several health related examples.
Important Dates & Deadlines
- October 9 - 11, 2013