การเปรียบเทียบการประมาณค่าพารามิเตอร์ของตัวแบบอนุกรมเวลาที่มีฤดูกาล / ศรีสุดา ทิมกระจ่าง = A comparison on estimation of parameters in seasonal time series model
The purpose of this research is to compare the parameter estimation methods in seasonal time series models. These methods are Unconditional Least Squares (ULS), Conditional Least Squares (CLS) and Maximum Likelihood Estimation (MLE). The time series models are ARIMA(0,0,0)(1,0,0)4, ARIMA(0,0,0)(0,0,1)4, ARIMA(0,0,0)(1,0,1)4 and ARIMA(0,0,0)(2,0,1)4. The sample sizes are 60, 80, 100 and 120 quarters. This study simulates and analyzes data by using R program. The experiment was repeated 500 times under each condition and the criterion of determination are the mean squared error (MSE) or the average of mean squared error. Results of the study are as follows:- The sample sizes effect to the efficiency of estimate parameter in all time series models. 1. For ARIMA(0,0,0)(1,0,0)4 model, the parameters and the sample sizes are small. ULS method is the best method to the minimum MSE. In case of, the parameters and the sample sizes are increase, MLE method is better than 2 methods. MLE is the best method if the parameters are large for any sample size. 2. For ARIMA(0,0,0)(0,0,1)4 and ARIMA(0,0,0)(1,0,1)4 models, the parameters and the sample sizes are small. ULS method is the best method to the minimum MSE. In case of, the parameters and the sample sizes are increase, CLS and MLE method are better than ULS method. The large parameters, MLE method will gives the minimum MSE for ARIMA(0,0,0)(0,0,1)4 model. 3. For ARIMA(0,0,0)(2,0,1)4 model, ULS method is the best estimation of parameter to the minimum, for every value of parameter and sample size.
