Automatic ARIMA Forecasting

For econometric discussions not necessarily related to EViews.

Moderators: EViews Gareth, EViews Moderator

shehramkhan
Posts: 2
Joined: Tue Oct 13, 2015 8:32 pm

Automatic ARIMA Forecasting

Postby shehramkhan » Tue Sep 18, 2018 8:36 am

Hi,

I ran automatic ARIMA test on our daily sales data which gave the best model. I tried doing forecasting based on the best model but the standard errors were too large. So, I changed the dependent variables from DLOG(Daily_Units_Sold) to just 'Daily_units_Sold' and the standard errors became normal. Is it because of the log differentiation that it caused the standard errors to inflate so much? If so, then is there a way to normalize the standard errors?

Thank you,

Shehram Sikander

startz
Non-normality and collinearity are NOT problems!
Posts: 3400
Joined: Wed Sep 17, 2008 2:25 pm

Re: Automatic ARIMA Forecasting

Postby startz » Tue Sep 18, 2018 8:39 am

In one case you are predicting the first difference and in the other you are predicting the level.

shehramkhan
Posts: 2
Joined: Tue Oct 13, 2015 8:32 pm

Re: Automatic ARIMA Forecasting

Postby shehramkhan » Tue Sep 18, 2018 8:43 am

I agree. However, does taking the dlog of the dependent variable affects the variance so much to the point that it is almost impossible to see the dependent variable in a normal graph. Should I instead go with other top 20 models mentioned in the automatic ARIMA output?

Capture.JPG
Capture.JPG (61.23 KiB) Viewed 1227 times


Return to “Econometric Discussions”

Who is online

Users browsing this forum: No registered users and 7 guests