Title :
AR processes with non-Gaussian asymmetric innovations
Author :
Bondon, Pascal ; Li Song
Author_Institution :
Univ. Paris-Sud, Gif-sur-Yvette, France
Abstract :
We consider the problem of modeling non-Gaussian correlated signals by autoregressive models with skew exponential power innovations. Generalized moments and maximum likelihood estimators of the parameters are proposed and large sample properties are established. Finite sample behavior of the estimators is studied via Monte Carlo simulations. An application to real data is considered.
Keywords :
Monte Carlo methods; autoregressive processes; maximum likelihood estimation; AR processes; Monte Carlo simulations; generalized moments; maximum likelihood estimators; non-Gaussian asymmetric innovations; non-Gaussian correlated signals; skew exponential power innovations; Biological system modeling; Covariance matrices; Data models; Maximum likelihood estimation; Noise; Technological innovation; Non-Gaussian; asymmetric distribution; autoregressive model; maximum likelihood estimation;
Conference_Titel :
Signal Processing Conference (EUSIPCO), 2013 Proceedings of the 21st European
Conference_Location :
Marrakech
