Page 1 of 1

HW programming question

Posted: Tue Jun 03, 2014 9:52 am
by leonardo
Hello all,
I have a question regarding E-Views and I would like to know if anyboady can help me with it.

I have ten years of quarterly broadband connections data for more than 100 countries, and I run on them the Holt-Winters’ three-parameter exponential smoothing method (multiplicative form) to forecast the number of connections for the next 12 quarters.

In the next quarter, once the actual data is available I have to re specify the three parameters for each data points using the actual data and based on the new data the result of forecast would be different.

How do I re specify this in an automated way, not having to run each process one by one?

Basically I need the syntax for HW smoothing and a simple program to re specify the value of three parameters automatically and make a new set of forecasts.

Thanks in advance

Re: HW programming question

Posted: Tue Jun 03, 2014 9:58 am
by EViews Gareth

Code: Select all

series.smooth
is the command to perform Holt-Winters exponential smoothing. More details on it can be found in the series section of the Object Reference pdf. Without knowing how you want to choose the parameters, it is difficult to tell you how to program it.

Re: HW programming question

Posted: Wed Jun 04, 2014 2:09 am
by leonardo
Gareth,
Thanks for the prompt reply. Returning to your message, I use a model like this smooth(m,.517,.03,.9) for each country based on data up to the first quarter of 2014.
To choose/adjust the current parameters I'd like to use the gap between mean absolute percentage error (MAPE) of current forecasts and the one which will be based on data up to the second quarter of 2014 once the real data are published. I was wondering if:
1- The gap between the two MAPEs of two forecasts can be used to re-specify my model?
2- How to re-specify the model using the above in an automated way foe whole lots of countries?

Thanks
Leo

Re: HW programming question

Posted: Wed Jun 04, 2014 7:53 am
by EViews Gareth
That's an econometric issue.

Re: HW programming question

Posted: Wed Jun 04, 2014 8:10 am
by leonardo
Yes you are nearly right in the sense that selection of MAPE, error terms etc to re-specifying the model is an econometric issue but the point 2 is a programming issue and you might be able to address that. Many thanks
Best
leo

Re: HW programming question

Posted: Wed Jun 04, 2014 8:24 am
by EViews Gareth
Once you've decided what you want to do, we can possibly help in figuring out how to do it. But we can't answer 2) until you've figured out 1).

Re: HW programming question

Posted: Wed Jun 04, 2014 9:06 am
by leonardo
Many thanks. I'll keep you posted about the selection criterion.
Best
Leo