Fisher's z Autoregressive Model: A Case Study on Forecasting Gross Regional Domestic Products at Constant Prices in East Kalimantan Province
Keywords:Fisher's z autoregressive model, fat tailed, GRDP at constant prices forecasting, Bayesian method, Stan
In this paper, we propose the Fisher's z Autoregressive (ZAR) model with its application. The ZAR model is an Autoregressive (AR) model whose error term has a Fisher's z distribution. The shape of the Fisher's z distribution, which can be skewed or symmetrical, makes this distribution more flexible in capturing fat tailed patterns in the data, compared to the Gaussian distribution. Furthermore, the ZAR model is used to forecast the Gross Regional Domestic Product (GRDP) at constant prices 2010 in East Kalimantan Province in the first quarter of 2022. As a comparison, in this study also forecasts real GRDP using the Gaussian-AR (GAR) model. The method used to estimate the parameters of the models are the Bayesian method using the Hamiltonian Monte Carlo (HMC) algorithm. This study also included the Stan code for fitting the ZAR model.