Asymptotic behavior of unstable INAR(\(p\)) processes (Q550155)

The authors investigate the asymptotic properties of integer-valued autoregressive model \(INAR(p)\) determined by the equation \[ X_k=\alpha_1\circ X_{k-1}+\cdots+\alpha_p\circ X_{k-p}+\varepsilon_k,k\in\mathbb N, \] where \(\alpha\circ X=\sum_{j=1}^X\xi_j\) if \(X>0\), \(\alpha\circ X=0\) if \(X>0\), \(\xi_j,\;j\in\mathbb N\), are i.i.d. Bernoulli random variables with mean \(\alpha\in[0,1]\), \(\varepsilon_k\) are i.i.d. non-negative integer-valued random variables with the mean \(\mu_{\varepsilon}\), and \(\alpha_1,\dots,\alpha_p\in[0,1]\). Under the assumption that the second moment of the innovation \(\varepsilon_k\) is finite the authors proved that the sequence of appropriately scaled random step functions \(X^n_t=X_{[nt]}/n\), \(t\in\mathbb R_+\), \(n\in\mathbb N\), formed from an unstable \(INAR(p)\) process (\(\alpha_1+\cdots+\alpha_p=1\)) converges weakly towards a squared Bessel process determined by the stochastic differential equation \(dX_t=\left( \mu_{\varepsilon}dt+\sqrt{\sigma^2_{\alpha}X_t^+}dW_t \right)/\varphi'(1)\), \(X_0=0\), \(t\in\mathbb R_+\), where \(\varphi'(1)=\alpha_1+2\alpha_2+\cdots+p\alpha_p>0\), \(\sigma^2_{\alpha}=\alpha_1(1-\alpha_1)+2\alpha_2+\cdots+p\alpha_p(1-\alpha_p)\), \(W_t,\;t\in\mathbb R_+\), is a standard Wiener process. This limit process is a continuous branching process also known as square-root process or Cox-Ingersoll-Ross process. This asymptotic behavior of unstable \(INAR(p)\) models is quite different from that of familiar (real-valued) unstable autoregressive processes of order \(p\) (\(AR(p)\) models). An application to Boston armed robberies data is presented.