Evaluación de modelos de pronóstico de series temporales para el Índice del mercado colombiano COLCAP

Abstract

The main objective of this research to focus on the selection and adjustment of two autoregressive models of the COLCAP index performance series, following the Box-Jenkins methodology or the ARIMA methodology. The ARMA (p, q) and ARMA (p, q) - GARCH (p, q) models are chosen and adjusted. The obtained results show that these series can be represented with an autoregressive model of order four. Besides, by the given the presence of heterocedasticity and the ARCH effect in the squared residuals of the model, the adjustment of the GARCH model (p, q) was performed. Finally, the forecast performance is evaluated by means of error measurement, in which the ARMA model presents a lower error value.