Search for: [Abstrakt = "The aim of the paper is to compare the forecasting potentials of two classes of Multiplicative Stochastic Factor – scalar BEKK \(MSF\-SBEKK\) models which differ in the type of latent process. In the first class, the innovations of a first order autoregressive structure for the natural logarithm of latent variables are log\-normal, whereas in the second class the innovations are inverted gamma distributed. The comparison of the models’ forecasting abilities by means of the predictive Bayes factor as well as the log predictive score and energy score were performed based on the Polish exchange rates. The author considered one\- to ten\-step\-ahead predictions during the period beginning from 3 September 2019 and ending on 2 September 2020, which covers the time of the crisis caused by COVID\-19. The author concluded that for most of the forecast horizons, the considered log\-normal innovations outperformed the inverted gamma ones"]
