Title: Computation of interval forecast for ARIMA models accounting for the uncertainty of parameters' estimates
Authors: Nikita Moiseev - Plekhanov Russian University of Economics (Russia) [presenting]
Nikolay Tikhomirov - Plekhanov Russian University of Economics (Russia)
Abstract: A numeric method is introduced for calculating the confidence interval for ARIMA type models, taking into account the uncertainty of parameters' estimation, what is especially important with a relatively short data frame. The proposed method is based on estimating the model parameters through the Yule-Walker system of equations, which uses autocorrelation coefficients of various orders. A method is presented for estimating the variance-covariance matrix for the autocorrelation coefficients with the subsequent application of the numerical method for estimating the variance of the regression function. In addition to the theoretical calculations, simulations are carried out to test the developed method in comparison with the traditional method of obtaining the interval forecast.