Estimating on sub sample, predicting for full sample

[MichaelS]

I wish to estimate expenditure on a sub-sample, and use the estimates to predict expenditure for the entire sample. So in the first stage, I regress:

proc reg data=InputData outest=estimates;

  model Exp = lnY lnY_2 &ivset ;

  where Exp ge  100;

  output out=Pred_Exp p=predexp;

run;

The problem is I then have to manually use the parameter estimates to predict expenditure for the cases with Exp<100. Is there a shorter way to do this than manually using the parameter estimates that I have saved to the dataset estimates?

[SteveDenham · Accepted solution]

In the input dataset, create a variable (call it Exp1).

data inputdata;

set inputdata;

if exp >= 100 then exp1=exp;

     else exp1=.;

run;

then run proc reg as follows:

proc reg data=InputData outest=estimates;

  model Exp1 = lnY lnY_2 &ivset ;

  output out=Pred_Exp p=predexp;

run;

The estimates dataset will have predicted values for all of the data.  This use of missing for the dependent variable is a good trick to know.

Steve Denham