Nonparametric Selections in Blocked Data: Application to Motor Vehicle Fatality Rate Data

Abstract

Selection procedures applicable to the blocked data were reviewed. When applied to the state traffic fatality rate data, rules based on ranked observations performed most poorly among the procedures reviewed. The data failed not only a normality test but also a test on the location parameter model assumption. We infer from the relative efficiency study that the rank procedures may not be satisfactory if the distribution underlying the data is not from a location family (or the like). Procedures to select the most probable multinomial category, applied to the blocked data, show a promising result as nonparametric methods.

KEY WORDS:

• Subset selection
• Block design
• Ranked data
• Selecting the most probable multinomial event
• Test of location model assumption
• Relative efficiency