B0672
Title: Mixtures of Kato-Jones distributions on the circle with an application to traffic count data
Authors: Kota Nagasaki - Tokyo Institute of Technology (Japan)
Shogo Kato - Institute of Statistical Mathematics (Japan) [presenting]
Wataru Nakanishi - Tokyo Institute of Technology (Japan)
Abstract: Kato-Jones distribution is a probability distribution on the circle that is unimodal and affords a wide range of skewness and kurtosis. Motivated by a multimodal skewed data set which appears in traffic engineering, mixtures of Kato-Jones distributions are considered. A key reparametrization is done to achieve the identifiability of the proposed mixtures. With this reparametrization, two methods for parameter estimation, namely, a modified method of moments and the maximum likelihood method, are presented. The modified method of moments estimation is relatively fast and provides the reasonable initial value of the algorithm for the maximum likelihood estimation. The maximum likelihood estimation can be done using the EM algorithm. These two methods are seen to be useful for fitting the proposed mixtures to the traffic counter data set of interest.
