11-19-2014 12:20 PM
I'm using the HPFDIAGNOSE Procedure and I've got a question about ouliers and input selection.
I'm searching for outliers (proc hpfdiagnose statement), and also diagnosing over a list of inputs (required=maybe). In the output I can see that outliers are identified, and are treated by using pulses, so I can say that I've got new "inputs" in my model (the pulses).
My question is: Are these "new inputs" (caused by the outliers) analized together with my inputs, (so that if there is collinearity, etc. the outlier pulse could probably not be included), or the selection of regressors is only made on the set of inputs (input statement, required=maybe) independently of the treatment of the outliers?
Thanks a lot in advance,
11-26-2014 04:08 PM
Details about detecting outliers can be found here: http://support.sas.com/documentation/cdl/en/etsug/67525/HTML/default/viewer.htm#etsug_arima_details3... - please note that outliers use residuals after ARIMAX components are estimated. Outliers must additionally improve AIC/SBC to be included by HPFDIAGNOSE.