Department of Mathematics

Speaker: Ramo Gencay
Title: Multi-Scale Tests for Serial Correlation
Affiliation: Simon Fraser University
Date: Thursday, March 21, 2013
Place and Time: Room 102, Love Building, 3:35-4:30 pm

Abstract. This paper introduces a new family of portmanteau tests for serial correlation. Using the wavelet transform, we decompose the variance of the underlying process into the variance of its low frequency and of its high frequency components and we design a variance ratio test of no serial correlation in the presence of dependence. Such decomposition can be carried out iteratively, each wavelet filter leading to a rich family of tests whose joint limiting null distribution is a multivariate normal. We illustrate the size and power properties of the proposed tests through Monte Carlo simulations.