Tests for Local Temporal Correlation of Two Non-Homogeneous Poisson Processes

Barry James and Kang James
Department of Mathematics and Statistics
University of Minnesota- Duluth

Wednesday, May 7th
3:30pm
MoosT 5-125
Minneapolis Campus

Abstract:

We study the local temporal correlation of two non-homogeneous Poisson processes, where the processes are the times of successive occurrences of two different events. Hugeback (2007) modeled two types of solar events, solar flares and coronal mass ejections, as non-homogenous Poisson processes and studied the correlation of the two processes in a (small) local window. She proposed parametric models for the distribution of relative time points and used likelihood-ratio tests and the bootstrap to test the significance of the correlation of the two processes. Our point of view is nonparametric. We derive a test statistic based on a small window of differences between all pairs from the two processes. The statistic is a U statistic, and we obtain its asymptotic distribution. An easier ad hoc test based on the binomial distribution will also be discussed. Simulation results and some applications of the model will be given.

A social tea will be held at 3:00 P.M. in A434 Mayo. All are Welcome.
For more details contact 612-624-4655 or see http://www.biostat.umn.edu/seminar_academic.html