A0568
Title: Two step estimations via the Dantzig selector for ergodic time series models
Authors: Kou Fujimori - Shinshu University (Japan) [presenting]
Koji Tsukuda - Kyushu University (Japan)
Abstract: The estimation problems for conditionally heteroskedastic ergodic time series models are considered with high-dimensional and sparse parameters with some nuisance parameters. The asymptotic behavior of the Dantzig selector is first established based on the least squares method to estimate the unknown parameter of interest. Then, using the Dantzig selector and a consistent estimator for the nuisance parameter, the asymptotically normal estimator is constructed for the non-zero components of the parameter of interest. Applications to order selection problems for integer-valued autoregressive models of large order are presented.
