Abstract:
|
Often when we wish to rank a set of populations by some metric, the true parameter values for each population are unknown, and hence so is the true overall ranking. We discuss recent work on constructing and visualizing a joint confidence region for the true unknown ranking, assuming that we have an estimate and its estimated standard error for each population. Through real-data examples, we demonstrate the practical consequences of acknowledging uncertainty in rankings; some considerations in how to visualize and communicate the joint confidence region for lay audiences; available R packages for calculations and graphs; and current efforts at the US Census Bureau towards implementing these ideas in user-facing data products.
|