SAS Forecast Studio - Scoring Existing ARIMAX Models on New Data (Independent Variables)

[DrSharma]

Hi,

I have built some ARIMA models based on the true drivers of the response variable.

I want to generate forecasts of the response based on new values of the predictors with same set of parameter estimates.

Please help me understand how to achieve this and how prepare the new dataset required.

Thanks very much in advance.

Regards

[udo_sas]

Hello -

If using Forecast Studio all you have to do is to append the new history to the data set which you used to build the models.

Then, when you open your project you will be prompted what you would like to do.

One of the options if to reuse existing models and existing parameter:

"Forecast: refresh the current forecast model, using the same parameter values"

That should do the trick.

You can accomplish the same by using HPFENGINE of course.

Thanks,

Udo

[udo_sas]

PS: by the way - using the scenario analysis feature of SAS Forecast Studio you will be able to change future values of your predictor variables interactively.

[DrSharma]

Hi Udo,

Actually, I had some 20 time series to model and project for next 6 months.

Initially I had kept 50 data points (Apr'08 till May'12) to train the model and next 12 data points (Jun'12 to May'13) for validation.

So, I had 62 data points in the dataset for each of the 20 series being analyzed.

After building the models I replaced the dataset with new updated one having the next 6 future values of the x variables with response

variable values as missing.

I chose the same option as you've mentioned "Forecast: refresh the current forecast model, using the same parameter values".

But I still obtained the same future forecast values as I saw before updating the dataset with new values of x variables.

I thought I was making some mistake. Could you please shed some light?

Thanks once again.

[udo_sas]

Hello -

Please excuse for asking the obvious, but were the x variables included to your winning models?

Also, when you say validation data points, are you talking about hold-out samples or out-of-sample data?

How different are your next x values from the previous values and what do the parameter estimates for the x values look like?

Maybe your expectation about the impact of the x variables does not hold true?

Thanks,

Udo

[DrSharma]

Hi Udo,

Yes, the winning models were having the the x variables included.

By validation data I meant the holdout sample.

The new x values are significantly different but within the expected range, look normal as the rest of the values.

The parameter estimates of the transfer model components take different combinations, sometimes with very small p-values, and sometime with very high values. But the p-values do matter here?

Thanks for your time

[udo_sas · Accepted solution]

Hello -

Thanks for clarification - the p-values should not matter, indeed.

At this point in time it is probably better to open a ticket with Technical Support, who can assist with tracking down this problem.

Thanks,

Udo