|
Activity Number:
|
146
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Monday, August 7, 2006 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Section on Statistics in Epidemiology
|
| Abstract - #307372 |
|
Title:
|
Surveillance of Occupational Drivers Using k Nearest Neighbor Methods on the Line
|
|
Author(s):
|
Svetla Slavova*+ and Terry Bunn and Dmitri Pavlov and Richard J. Kryscio
|
|
Companies:
|
University of Kentucky and University of Kentucky and Pfizer Inc. and University of Kentucky
|
|
Address:
|
KY Injury Prevention and Research Center and Department of Statistics, Lexington, KY, 40504,
|
|
Keywords:
|
case-control ; density ratio ; k nearest neighbor ; parallel implementation ; occupational driving ; surveillance
|
|
Abstract:
|
In the surveillance of non rare events on the line the ratio of the densities of cases to controls is used to define a local cluster whenever this ratio exceeds a predetermined constant. A nonparametric k Nearest Neighbor method (kNNM) is used to estimate this ratio since with large samples it yields simple to compute confidence intervals for the ratio and straightforward edge corrections. Numerical simulations confirm these asymptotic properties. A parallel implementation of the kNNM is proposed which does not compromise these properties. Applications to two large Kentucky Occupational Safety and Health Surveillance databases show that out of state drivers have significantly higher accident risk past age 50 with opposite results for the youngest drivers and that occupational drivers have a higher accident risk during the early morning hours.
|
