Heng Chen, The Bank of Canada
Project Description/Abstract
We consider statistical inference of jumps in nonparametric regression models with long memory noise. A test statistic is proposed for the presence of jumps based on a robust estimator of the variance of the wavelet coefficients. The sequential applications of tests allow us to estimate the number of jumps and their locations. Compared to the existing inference procedure whose test statistics converges to the extreme value distribution very slowly, ours processes more accurate finite sample performance derived from the asymptotic normality of our test statistics.
Co-author
Mototsugu Shintani, The University of Tokyo