summarized in Tables 5.1 and 5.2. From these results, we can find the criteria of MAIC 
that give us the correct values of the order except in a few cases. Estimation of parame- 
ters @}- - - @,3b,- - - b,, and az is rather reasonable in these examples. Tables 5.1 and 
5.2 list the AR(n) model fitted to the original ARMA or MA models. It is interesting that 
the values of N obtained by the MAIC method are somewhat larger than n or m of the 
original processes. These models are the AR models which approximate most closely the 
original ARMA or MA models. We can say these examples gives us good proof that the 
model fitting method, supported by order determination through AIC criteria, is a very 
reasonable and powerful method for analyzing the linear stochastic processes. 
199 
