scientific article; zbMATH DE number 3932251

From MaRDI portal
Publication:3704776





We consider the problem of estimating autoregression coefficients in a situation in which the distribution of the noise is incompletely known. Consideration is given to recurrent estimates (of the stochastic approximation type) and nonrecurrent estimates (of the maximum likelihood type), including arbitrary nonlinear transformation of the residuals. Asymptotic properties of such estimates are given. For the cases in which the noise distribution is completely known (or partially known - with accuracy up to membership in some class of distributions), optimal estimates are constructed. It is shown that, under broad assumptions though not always, such estimates are attained by the method of least squares, which proves to be stable under variation of the noise distribution in an autoregressive (as opposed to a regression) situation.











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 Q3704776)