STARMA Models Estimation with Kalman Filter: The Case of Regional Bank Deposits

Abstract

In this study, STARMA Models' performances and advantages on analysis of variables based on time and space are addressed. Even though the history of STARMA Models starts from 1980's, these models remained in the shadows because of both lack of variable sets and its estimation difficulties. When the literature is examined, it's seen that STARMA Models can be estimated by linear and non-linear estimators. It's seen that the non-linear estimators are producing more efficient results because of the variable's structure. For this reason, in this study, Kalman Filters and maximum likelihood estimator has been used. The calculation of spatial weight matrix is done with software Kure which is developing by our team. As the case study, regional deposits of commercial banks operating in Turkey were analysed. Statistically significant and robust results revealed that STARMA Model has high estimation performance. (C) 2015 The Authors. Published by Elsevier Ltd.