Comparison of Model Selection Criteria for Models Including Trend and Seasonal Components in Econometric Time Series

The paper aims to compare commonly used model selection criteria in time series modeling, such as Adjusted R2, log-likelihood, Akaike Information Criterion (AIC), Schwarz Information Criterion (SIC), Hannan-Quinn (HQ) Information Criterion, and Mean Squared Error (MSE). In this context, for an additive time series, data was produced in different sample sizes from n=60 to n=500 from (17) different stationary stochastic processes, including constant, trend, seasonal and irregular components. Each production was repeated 10000 times and the criteria were calculated. For very large sample sizes, the HQ information criterion provides the best results for all types of time series models. It was observed that log-likelihood performed poorly in almost all models. It has been found that "Adjusted R2" is the best option for models with sample sizes less than 120, and "AIC" criterion is the best option for choosing the right model as the sample size increased.