Asymptotic theory of estimation of parameters in autoregressive models under general set-up of the roots

In this paper the asymptotic distribution of the least square estimator of the oarameters of a pth order autoreqressive process with stationary φ -mixing error variables having uniformly bounded fourth order moments has been considered. (That the error variables are also Gaussian has been assumed for the explosive case). A comoo-nent-wise break up of the process with respect to the roots has been made and the limiting distribution of each comoonent obtained by using the invariance principle of φ-mixing processes. The main aim of this paper is to reflect on the explosive case where a refinement of Anderson's technique (1959) has been made.