4/18/2014

Post date: Apr 21, 2014 3:13:32 PM

12:00-1:15pm Biostatistics Library, Lunch event

Drs. Paul Albert and Dean Follmann: "Statisticians' career at NIH"

1:30-2:50pm Room E3607, SLAM Seminar

Speaker: Dr. Paul Albert, Biostatistics and Bioinformatics Branch, NICHD/NIH

Title: The interface between statistical research and teenage driving: what statistics can teach us about how our kids drive and visa-versa.

Abstract

Motor vehicle crashes are the leading cause of death for U.S. teens. Investigators at the NICHD are conducted a series of longitudinal studies that closely monitor teen driving using in-vehicle data recording systems. Specifically, instrumentation is installed in their cars that include accelerometers, video cameras, a global positioning system, front radar and a lane tracker. Of interest is examining how teenagers drive over the first 18 months of licensure and whether g-force events can be used to predict accidents. These studies have led to interesting collaborations between statistical and behavioral investigators. This talk will discuss new statistical methodology that was developed to address analytic challenges encounter by the study team. One analytic challenge concerned inference on trip-level g-force events that are measured on each trip made during the 18 month period. We discuss random effects and marginal approaches for analyzing these data accounting for the large number of trips on each subject (thousands), between-subject variation, over-dispersion, and serial correlation. Another analytic challenge involved using prior g-force events to predict subsequent crashes. We develop latent variable and hidden-Markov modeling approach that can directly be used for risk prediction in this setting. The highlights of each approach are presented along with interesting data from the Naturalistic Driving Study of the NICHD.