January 1, 2015

Two R Packages for Sensitivity Analysis in Observational Studies

By Paul Rosenbaum

packpaperos_published

Two R packages for sensitivity analysis in observational studies are described. Package sensitivitymw is for matched pairs with one treated subject and one control, or matched sets with one treated subject and a fixed number, K  2, of controls. Package sensitivitymv is for matched sets with variable numbers of controls. The packages offer conventional statistics, such as the permutational t-test and M-statistics using Huber’s weights, but they also offer less familiar test statistics that have higher power in sensitivity analyses. The packages provide several tools useful in sensitivity analyses, such as an aid, amplify, to the interpretation of the value of the sensitivity parameter, and a device for combining evidence from several independent sensitivity analyses, truncatedP, for instance, several evidence factors or several subgroups.


Software Descriptions