A Non-Parametric Application for Exploratory Multivariate Data Analysis Using Infant Mortality Rates

Ann Von Holle, University of Pennsylvania
Chirayath Suchindran, University of North Carolina at Chapel Hill

Simple exploratory analyses often produce mean and standard deviation estimates. This presentation offers an alternative approach with the primary objective being the application of a moment-free method to characterize multivariate data including county-level North Carolina infant mortality rates from 1989-2000. This method involves simplicial depths, originated from probabilistic geometry in the past 15 years, and its use applies to a wide range of other multivariate data regardless of distribution type. This alternative, non-parametric approach to describe data yields equivalents to center, dispersion and peakedness measures of the data as well as means to identify departures from a distribution. In this case, potential distributional differences occurring over time will be examined. This application concludes with an extension to explore infant mortality differentials by race and education values, showing more disperse rates over time given a multivariate structure.

  See paper

Presented in Poster Session 6: Applied Demography, Methods, Migration, Labor and Education, Gender, and Race and Ethnicity