Application of one-dimension STS-distribution for modelling magnitudes of stock indexes

Abstract

Modified method STS-GARCH(1,1) has been considered. Modification consisted in rejection of the statement on normal low of logarithm distribution of time series day increment and in their application for the description of Smoothly Truncated a-Stable (STS)-distribution (smoothly abridged a-stable). The method parameters were found by the technique of maximum likelihood. Statistic investigation of the suggested algorithm accuracy was carried out and decrease of autocorrelation in data structure used for the analysis was shown. The method was used to predict share prices of lag 5.