12-05-2011 03:35 PM
Let's say I have a dataset of 1000 observation.
I want to construct a model using the first 900 observation, and to build predictions for the 901th -1000th observation-in orde to test my model and compute the residual.
How do I get predictions to the rest of my dataset?
12-05-2011 04:00 PM
Take a look at proc score http://support.sas.com/documentation/cdl/en/statug/63033/HTML/default/viewer.htm#statug_score_sect00...
It will accept the model that you can output from proc reg.
However, you probably don't want to take the first 900 observations but, rather, construct to datasets one containing a random selection of 900 records and the other the remaining 100 records.
12-06-2011 11:07 AM
Thanks. my serie is a Time serie- and for that reason I don't chose the 100 records randomly.
I want to compare classical regression results with reg-arima (dependent variables+arima for residuals)
I'll try to raed about the score procedure.