دراسة نموذج ARMA-GARCH عندما يتبع الخطأ توزيع Laplace ومقارنته مع التوزيع الطبيعي

Abstract

     ARMA-GARCH model study  on  of the recent studies in the field of time series, and that the most important characteristic of these models to both conditional mean and conditional variance depends on the past. And most of these studies used three continuous distributions of conditional error is Normal ,Student – t and General Error distributions ,and discrete type are the Poisson and the negative binomial distributions.

     In our research, one continuous conditional error distribution is suggested, include Laplace distribution. And then study the distribution proposed theoretically, empirically and practically. Thus the goal of the thesis is to study the time series when observations are real values and follows the random error of GARCH model in continuous distributions.

    The results of the experimental side in case of continuous distributions had been reached that forms where the random error follows Laplace distribution was preferable compared with the normal distribution.

In practical side examined application took a representative sample of daily price in Japanese stock market that was distributed as Laplace distribution and also suffers from the problem of heteroscedasticity of the conditional variance of error, and remove the influence problem matching AR(1)-ARCH(2) model