Two Types of Gaussian Processes and their Application to Statistical Estimations for Non-ergodic Vasicek Model

Yong Chen, Ying Li, Yanping Lu

Published 2023-10-02Version 1

For the non-ergodic Vasicek model driven by one of two types of Gaussian process $(G_t)_{t\in[0,T]}$, we obtain the joint asymptotic distribution of the estimations of the three parameters $ \theta,\mu$ and $\alpha=\mu\theta$ of the model. A novel contribution is the identification of three different types of scaling in the asymptotic behavior of the estimator $\hat{\theta}_T$ of $\theta$. The proof is based on a slight modification of the idea of Es-Sebaiy(2021). The two types of Gaussian process $(G)$ satisfy that either its first-order partial derivative of the covariance function or the difference between its covariance function and that of the fractional Brownian motion $(B^H)_{t\in[0,T]}$ is a normalized bounded variation function. There are ten Gaussian processes with nonstationary increments in the literature that belong to this category, and nine of them can be used as driven noise for the Vasicek model to obtain the joint asymptotic distribution. The same result concerning the non-ergodic Ornstein Uhlenbeck process is a by-product.