The Univesity of Liverpool
Department of Mathematical Sciences
Division of Statistics and Probability
& Department of Economics
SEMINAR
Testing for Neglected Nonlinearity in Long Memory Models
Richard T. Baillie Michigan State University and Queen Mary, University of
London
Wednesday, 9th March 2005, 2pm
The Whittaker Room (211)
Abstract:
The paper discusses some tests for the presence of general non-linear
components in time series processes in addition to a fractionally
integrated, long memory component.
The tests are based on artificial neural network structures and also on
high order Taylor series expansions; and do not restrict the form of
non-linearity being investigated. All the tests require consistent
estimation of the long memory parameter and we discuss three alternatives:
local Whittle, time domain non-linear approximate MLE and the Fox Taqqu
estimator in the frequency domain.
Some detailed simulation evidence is presented on the performance of the
estimators of the long memory parameter and also the size and power of the
suggested test statistics under various alternative Exponential Smooth
Transition AutoRegressive (ESTAR) data generating processes.
The paper also applies the methodology to various economic and financial
time series problems; including the debate on whether monetary policy has
caused structural breaks and non linear features as opposed to long memory
in inflation. We also consider real exchange rates and the debate over mean
reversion to purchasing power parity. Also, the relatively new formulation
of Realized Volatility (RV) is investigated in currency and commodity
options markets.
We also discuss interpretation of the empirical work and some remaining
research issues.
Following the talk, tea and biscuits will be available in Room 304
ALL WELCOME
********************************************************
Ingrid Harper
University of Liverpool
Department of Mathematical Sciences
Division of Statistics and Probability
Peach Street
Liverpool L69 7ZL
Tel: 0151 794 4751
Fax: 0151 794 4754
********************************************************
|