scientific article; zbMATH DE number 1069579
A vector \(X_t\) generated by the Markov scheme \(X_t= PX_{t-1}+ \varepsilon_t\) will be called a `positive autoregressive dependent series' if \(P>0\), so that all elements of \(P\) are positive. Here \(\varepsilon_t\) is a zero mean white noise vector, so that \(\text{cov} (\varepsilon_{it}, \varepsilon_{jt})=0\), \(s\neq t\), all \(i,j\), and it is assumed that the distribution of \(\varepsilon_t\) does not depend on \(X_{t-j}\), \(j\geq 0\). A Markov process generated by this way with \(P>0\) may well be considered to be a reasonable model in many situations, such as when modelling a group of interest rates or exchange rates or highly related commodity prices or stock prices. A Bayesian may want to impose priors on \(P\) making it positive. The results found here indicate that in the long run a very simple structure arises from such a model and, if the dominant eigenvalue of \(P\) is \(\geq 1\), very special forms of cointegration arise. The fact that the processes are also associated allows further of their properties to be derived. For the entire collection see [Zbl 0846.00010].
This page was built for publication:
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4356540)
