## Robustness, Sensitivity, Reanalysis

Suppose famed researchers Steve and Jake are in the following situation:
• Jake has designed and carried out an experiment using prior parameters and

• Steve has prior parameters and

• Steve can
• design and conduct a new experiment, or
• acquire and re-analyze Jake's data using Steve's own prior parameters

### Goal

Assess the loss of efficiency to Steve of re-analyzing Jake's data instead of conducting his own experiment.

### Approach

1. Generate the allocation rule for each method (OS, BF, EA, HP, MP) using Jake's prior, but compute the risk using Steve's prior (i.e., Steve reuses Jake's data)
2. Generate and evalatuate the allocation rule for each method using Steve's prior (i.e., Steve conducts his own experiment)
3. Define the relative efficiency of each method to be the ratio of the risk of that method from (2) to the risk of the method from (1).

Note that an extremely large number of computations will be performed, since this will be done for many methods and many priors.

Robustness Results Outline Further Results