Non-asymptotic Confidence Estimation of the Autoregressive Parameter in ARMA(1,q) Model

Sergey Vorobeychikov, Andrey Pupkov
The article deals with the problem of confidence estimation of the autoregressive parameter of the ARMA(1, q) process in a non-asymptotic statement. It is assumed that the noise of the process is Gaussian with an unknown variance. The problem is solved by using a sequential modification of the Yule-Walker estimate.