University of Helsinki
Joint work with Pentti Saikkonen
PDF Document:
https://blogs.helsinki.fi/lkvaisan/files/2010/08/ReliableResiduals.pdf
Abstract:
We generalize and apply quantile residuals to multivariate nonlinear time series models for which conventional residuals are unreliable. We formulate a general framework of obtaining misspeciĀ
cation tests that also allows non-ergodic data and takes the effect of parameter estimation properly into account. Computationally simple tests developed to detect serial correlation, conditional heteroscedasticity,
and non-normality in quantile residuals illustrate the usefulness of our approach. Our tests are generalizations of previous tests based on moments of conventional residuals and the Lagrange Multiplier principle. We apply the developed tests to exchange rate series. In simulations our tests show good size and power properties.
An earlier version of the paper: HECER Discussion Paper No 247 / December 2008. PDF Document: http://ethesis.helsinki.fi/julkaisut/eri/hecer/disc/247/
Download estimation codes for GAUSS:
How to install Code for computing the tests Example code for estimating a multivariate GARCH model Dataset for the example code