Professor Carlos Velasco, University Carlos III

"Estimation of time series models by distance covariances"
Friday, 20 October. 3 pm
Room 411, Boyd Orr Building

Abstract

We develop estimates of time series models minimizing the distance autocovariances of residuals. These generalized autocovariances measure the dynamic dependence of time series described by integrals of the joint characteristic function. When the integrals use a nonintegrable weighting function with poles at the axis, the distance covariance has desirable properties, including characterization of serial independence, but the asymptotic analysis of the parameter estimates departs from the case with regular weights with finite moments. In this paper we concentrate on iterated estimates that use approximations to the score and Hessian of distances characterizing the serial independence and martingale difference properties of residuals with nonintegrable weighting. These measures are critical for identification of possibly noncausal and/or noninvertible models. We show that the new estimates are asymptotically normal distributed and investigate their performance in finite samples, showing that are competitive against optimally weighted estimates with maximum likelihood efficiency.

Bio

Carlos Velasco is Professor of Economics at University Carlos III de Madrid. His research fields are dynamic econometric models and economic time series analysis with applications in macroeconomics and finance.

For further information, please contact business-school-research@glasgow.ac.uk

We foster a positive and productive environment for seminars through our Code of conduct.

First published: 14 September 2023

<< 2023


Related links: