An integer-valued threshold autoregressive process based on negative binomial thinning (Q1785821): Difference between revisions

The authors study an integer-valued threshold autoregressive process based on negative binomial thinning operator (see [\textit{M. M. Ristić} et al., J. Stat. Plann. Inference 139, No. 7, 2218--2226 (2009; Zbl 1160.62083)]) which contains geometric counting series. First, they study some basic probabilistic and statistic properties for the first-order integer valued threshold autoregressive process NBTINAR(1) -- Propositions 2.1, 2.2 and 2.3. For parameters estimation they use two methods. First, the conditional least square (CLS) estimation procedure and study the asymptotic behavior of the estimator -- Theorem 3.1. Second, the conditional maximum likelihood (CLM) estimation method and establish consistency and the asymptotic distribution of the CLM-estimators -- Theorem 3.3 and Theorem 3.4. Their results, seem to extend some results given by \textit{M. M. Ristić} et al. [Commun. Stat., Theory Methods 41, No. 4--6, 606--618 (2012; Zbl 1237.62125)], \textit{M. Monteiro} et al. [Commun. Stat., Theory Methods 41, No. 15, 2717--2737 (2012; Zbl 1270.62122)], \textit{H. Zhang} et al. [J. Stat. Plann. Inference 140, No. 3, 667--683 (2010; Zbl 1177.62110)] or \textit{H.-Y. Kim} and \textit{Y. Park} [Stat. Pap. 49, No. 3, 485--502 (2008; Zbl 1148.62074)]. Some simulation and numerical results of the estimates and an application to a real data set are presented.