A0170
Title: ADCINAR(1) process and bias-correction of some estimators
Authors: Xiaoqiang Zeng - Hokkaido University (Japan) [presenting]
Yoshihide Kakizawa - Hokkaido University (Japan)
Abstract: The analysis of count time series is an emerging field. During the last four decades, there has been substantial progress on nonnegative integer-valued autoregressive (INAR) type models via the so-called binomial thinning operator (and its variant). The third and fourth auto cumulant (equivalently, central auto moment) functions are derived for the alternative dependent counting nonnegative INAR process of the first-order (ADCINAR(1)). A bias correction using a sample fourth auto moment function is also developed for the commonly used estimators; the Yule-Walker estimator and the conditional least squares estimator. The proposed bias correction, available without computing the closed-form expression for asymptotic expansions of the biases of these two estimators, is practically helpful since the bias formula for the case of the ADCINAR(1) process turns out to be rather complicated.
