A Review of Peter Lewis' Contributions to Nonlinear Time-Series Modeling

• Activity Number: 115
• Type: Topic Contributed
• Date/Time: Monday, July 30, 2012 : 8:30 AM to 10:20 AM
• Sponsor: Section on Statistics in Defense and National Security
• Author(s): Bonnie Ray*+
• Companies: IBM T. J. Watson Research Center
• Keywords: regression spline ; nonlinear ; periodic

Abstract

This talk will focus on the work of Peter Lewis from the late 1980's to his retirement from the Naval Postgraduate School in 1998, describing his work on nonlinear time series analysis based on regression spline methodology. The research was motivated by the work of Jerome Friedman at Stanford University in the mid-1980s on Multivariate Adaptive Regression Splines (MARS) for modeling nonlinear regression functions. Peter was one of the first to recognize the applicability of MARS for modeling nonlinear time series analysis, and proceeded to publish a series of important papers applying the methodology in the time series context, with extensions to allow for categorical predictors, seasonality, and periodic autoregressive behavior. I will review the work, share stories from my collaborations with him over the period, and show how Peter's research has continued to influence time series analysts in hydrology, climatology, finance, and beyond.