Erik Lindström,
Centre for Mathematical Sciences, Lund University

Non-linear Portmanteau Tests

Abstract

We derive Portmanteau tests for a large class of non-linear time series
models. This is done by replacing the linear measure of dependence with
a more general concept. Asymptotic distributions for the test statistics
are derived.

The methodology is computationally inexpensive, yet at least as flexible 
than most similar tests. All tests are also extended to heteroscedastic
data.