|
This paper is concerned with the distributional properties of a
median unbiased estimator of ARCH(0,1) coefficient. The exact
distribution of the estimator can be easily derived, however its
practical calculations are too heavy to implement, even though the
middle range of sample sizes. Since the estimator is shown to have
asymptotic normality, asymptotic expansions for the distribution
and the percentiles of the estimator are derived as the
refinements. Accuracies of expansion formulas are evaluated
numerically, and the results of which show that we can
effectively use the expansion as a fine approximation of the
distribution with rapid calculations.
Derived expansion are applied to testing hypothesis
of stationarity, and an implementation for a real data set is
illustrated. |
|
Keywords:ARCH(0,1) process,Edgeworth |
|