Journal of Statistical and Econometric Methods

Application of residual analysis in time series model selection

  • Pdf Icon [ Download ]
  • Times downloaded: 438
  • Abstract

    In this study, five criteria of residual analysis in time series modelling and forecasting are evaluated using three study variables namely, Nigeria’s Gross Domestic Product (GDP), Total Debts Accumulation (TDA) and Rate of Inflation (INFL). Considering five Auto Regressive Integrated Moving Average (ARIMA) specifications each for GDP and TDA and four ARIMA specifications for INFL, it was observed that four of the five criteria selected ARIMA(2,2,2) for the GDP I(2) while all the five criteria selected ARIMA(2,2,3) for TDA I(2) process. ARIMA(1,0,2) was also selected by all the criteria for INFL I(0) process. It is observed here that there is no particular criterion that clearly dominate others in the search for the “best” model specification and this suggests that modellers should consider the use of more than one criterion in model selection, especially when the family of ARIMA(p,d,q) models are of interest.