Towards a general theory for nonlinear locally stationary processes (Q1740517): Difference between revisions

After an almost complete solution given to the problem of statationary processes, the attention was focused on the study of nonstationary processes. Due to the large variety of nonstationary processes, concrete and special classes were studied. In this paper, the special class of locally stationary processes is investigated based on the stationary approximation and the stationary derivative. This study is organized in five main sections, presenting an introduction into the topic in the first one. In the second section, the stationary approximation is used to prove global and local laws of large numbers and a central limit theorem which holds under a stationary assumption, a dependence measure assumption and a localizing kernel assumption. In the third section, it is proved almost sure and a uniform Taylor expansion of the considered nonstationary process and a kind of differential calculus is obtained for locally stationary processes. Also some results on the spectrum and the empirical distribution function are obtained. In the fourth section, it is shown that a general Markov-structured locally stationary process fulfills the stationary approximation assumption and dependence measure assumption and, consequently, all obtained results in the previous two sections can be applied for such processes. In the fifth section, the asymptotic properties of maximum likelihood estimates for parameter curves of the considered locally stationary models are investigated using the asymptotic results and the obtained differential calculus. The paper ends with some examples and concluding remarks.