Abstract:

The design of observational studies is just as important as the design of randomised controlled experiments. Many observational studies now use propensity scores to make the treatment and control groups comparable. Although propensity scores remove the effect of measured covariates, they do not remove bias due to unmeasured variables. An essential aspect of the analysis and reporting of an observational study is to carry out a sensitivity analysis which determines the magnitude of the bias that would be needed to alter the conclusions of the study.

This talk summarises an R package, ObsSensitivity, that assists researchers to determine the appropriate sample size for an observational study. The software tool is developed as an R Commander Plug-in, giving the advantages of R but with a convenient and easily learnt environment. Demonstration of the use of the software with examples of actual observational studies is given.

Keywords:

Observational Studies, Power, Sensitivity Analysis, Package for Power Analysis.