Speaker
Dr. Thomas Fisher, Professor of Statistics, Department of Statistics, Miami University, Elected Member of ISI
Title
Statistics Seminar Series
Subtitle
Changepoints in non-Gaussian time series
Physical Location
Allen Hall 14
Abstract:
We present a method to identify multiple changepoints in non-Gaussian (discrete or continuous) autocorrelated time series. A transformation technique is used to derive the likelihood of a non-Gaussian series with an assumed latent autocorrelated Gaussian series, This enables penalized likelihood methods to handle changepoint identification and this methodology can be adapted for other forms of statistical modeling (e.g., linear modeling ideas with underlying autocorrelation). Here we present the results when the marginal distribution is a continuous proportion, modeled via a Beta distribution, and a multivariate series of counts, modeled with a multivariate Poisson.
Note:
Contact Prof. JZ at jzhang@math.msstate.edu for additional information.