Consistency Results for Stationary Autoregressive Processes with\n Constrained Coefficients

2017/06/08 by Alessio Sancetta, Sancetta, Alessio
Mathematics · #Statistical Methods and Inference

paper · pdf · doi:10.48550/arxiv.1706.02492

Abstract

We consider stationary autoregressive processes with coefficients restricted\nto an ellipsoid, which includes autoregressive processes with absolutely\nsummable coefficients. We provide consistency results under different norms for\nthe estimation of such processes using constrained and penalized estimators. As\nan application we show some weak form of universal consistency. Simulations\nshow that directly including the constraint in the estimation can lead to more\nrobust results.\n

Related

• Generalization error bounds for stationary autoregressive models
• A Theoretical Analysis of the Stationarity of an Unrestricted Autoregression Process
• Consistency and asymptotic normality in a class of nearly unstable processes
• Root-n consistent estimation of the marginal density in some time series models
• The Estimation of the Order of an ARMA Process