A Nonparametric Method for the Estimation of All Parameters in the Joinpoint Two Regime Regression Model

• Activity Number: 277
• Type: Contributed
• Date/Time: Tuesday, August 4, 2009 : 8:30 AM to 10:20 AM
• Sponsor: Biometrics Section
• Author(s): Lorrie L. Hoffman*+ and Steve Clark and Greg Knofczynski and Jaree Hudson and Heather King and Alfreda Rogers
• Companies: Armstrong Atlantic State University and SeaWorld and Armstrong Atlantic State University and Armstrong Atlantic State University and Armstrong Atlantic State University and Armstrong Atlantic State University
• Keywords: maximum likelihood ; Hinde index ; moment match ; Rao-Cramer ; change point ; bandwidth

Abstract

Hinde indices explain closeness of mother animals to their newborns---a proximity affinity calculated over time from data that includes number of approaches (actions reducing the distance between the pair) and leaves (those separating them) initiated by each animal in the pair. The underlying model is one of a decreasing index (showing less mother-infant dependency) that stabilizes at some point, reaching a steady state (maturation of the animal pair). Under this assumption, a two-regime regression model is feasible to uncover the parameters: an original proximity propensity, a rate of decrease to, level of affinity at, and time to reach steady state, and variability. Exploring a whale data set via the Quandt method (1958) reveals short-comings of this approach, leading to a discussion of more recent methods along with a moment-matching/nonparametric algorithm developed by the authors.