We consider the estimation of the parameters $s = (\nu, \alpha_1, \alpha_2, \cdots, \alpha_T)$ of a cumulative INAR($\infty$) process based on finite observations under the assumption $\sum_{k=1}^T \alpha_k < 1$ and $\sum_{k=1}^T\alpha_k^2<\frac12$. The parameter space is modeled as a Euclidean space $\mathfrak{l}^2$, with an inner product defined for pairs of parameter vectors. The primary goal is to estimate the intensity function $\Phi_s(t)$, which represents the expected value of the process at time $t$. We introduce a Least-Squares Contrast $\gamma_T(f)$, which measures the distance between the intensity function $\Phi_f(t)$ and the true intensity $\Phi_s(t)$. We further show that the contrast function $\gamma_T(f)$ can be used to estimate the parameters effectively, with an associated metric derived from a quadratic form. The analysis involves deriving upper and lower bounds for the expected values of the process and its square, leading to conditions under which the estimators are consistent. We also provide a bound on the variance of the estimators to ensure their asymptotic reliability.
翻译:暂无翻译