Forecasting With Underlying Factors

[kaodubela]

Hi Everyone.

I am interested to know the criteria SAS VA uses to identify the significant underlying factors that contribute to a forecast. I have not seen it mentioned in any of the documentation I have read online.

Thanks.

[Sam_SAS]

Hello,

The documentation is deliberately vague on this type of detail for several reasons.

My understanding is that VA creates a large number of forecasts using different algorithms, parameters, and underlying factors, and then it picks the combination of these that gives the "best" result. Any underlying factors that were included in the forecast that produced the best result are noted as having contributed to the forecast.

What I am not sure about (and might not be free to share) is how VA decides which results are the best.

Someone else might be able to get into more detail if that is what you want.

Sam

[kaodubela]

Thanks Sam.  I used VA in my research and I am currently writing a paper where I need to mention the criteria (in general terms) for choosing an underlying factor as significant. I just want to know the rationale for purpose of explanation in my paper. For example, is it based on correlation with the dependent variable, etc.

Thanks.

[Sam_SAS]

I'll see if the dev team can share any more details.

[kaodubela]

Thanks Sam!

[Sam_SAS]

My understanding is that VA's forecasting uses much of the same code as the HPFDIAGNOSE procedure to select a model. The underlying factors are considered as inputs (see the INPUT option).

I don't understand the nuts and bolts of forecasting enough to be able to give you a better answer.

[kaodubela]

Thanks Sam.