ARIMA model forecast for nonstationary data

hmhsinghmhsing Member Posts: 24 Maven
edited September 23 in Help
Hello,

I am using ARIMA(3,1,3) to forecast the S&P 500 close index. The (p,d,q) are determined from Optimize Parameters (Grid). However, the forecast values look strange to me (way too big). If I change to ARIMA(3,0,3) it turns out fine. Does this mean when d is not zero I need to do some other way to forecast?   





The rmp and data files are attached. Thanks for anyone's help! 

Tagged:
Sign In or Register to comment.