Comparison of spatial weight for consumer price index modeling in East Java using generalized space-time autoregressive

Abstract

Consumer Price Index (CPI) is one of the important economic indicators that can provide information regarding the development of the price of goods or services paid by consumers. The rate of change CPI that occurs will reflect the purchasing power of the money people use to meet their daily needs. CPI data collected at several years and cities can be modeled using Generalized Space-Time Autoregressive (GSTAR). GSTAR model has two parameters, namely time series and spatial parameter. The GSTAR parameters values in each location have different values so that this model is more suitable for modeling in several different locations. Determination of spatial weights that indicate the data relationship between locations has an important issue in GSTAR. The purpose of this research is to obtain the best spatial weight of three compared spatial weights in GSTAR models namely normalized cross-correlation, uniform, and inverse distance. CPI data in seven cities in East Java from January 2014 until August 2021 are used in this research. The result shows that normalized cross correlation weight has the smallest value AIC and MSE, respectively 1155,72 and 0,582. This indicates that normalized cross-correlation is the best spatial weight for East Java CPI using GSTAR.