Premium
Asymptotic Robustness of Prediction Intervals of Arima Models to Deviations From Normality
Author(s) -
Heuts R. M. J.
Publication year - 1981
Publication title -
australian journal of statistics
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.434
H-Index - 41
eISSN - 1467-842X
pISSN - 0004-9581
DOI - 10.1111/j.1467-842x.1981.tb00792.x
Subject(s) - autoregressive integrated moving average , kurtosis , asymptotic distribution , normality , mathematics , econometrics , gaussian , statistics , robustness (evolution) , local asymptotic normality , time series , estimator , physics , quantum mechanics , biochemistry , chemistry , gene
Summary ARIMA processes are extremely important in applied statistics. However, most theory is based on Gaussian stochastic processes, while it was pointed out by Mandelbrot (1965), Fama (1965), Heuts (1978) and others, that for many financial data the distribution of the shocks in the ARIMA schemes appears leptokurtic. In this paper we shall investigate the implications of some asymptotic results, not assuming normality, for the construction of prediction intervals of future observations in an ARIMA scheme.