The RSCORE macro no longer works!  Therefore, one must run
his/her own SAS code to obtain  R^2  for the validation (or
test) subset.  Detailed instructions are below.


1. Issue the following commands to SAS (with appropriate
   substitutions for "U:\").

LIBNAME GFLib 'U:\dm\sasdata' ; 

LIBNAME RCLib 'U:\dm\RC' ;


2. Calculate the sample mean of the response variable for
   the validation (or test) subset.  The example code below
   computes the sample mean of forced expiratory volume for
   the validation subset in {fevvalid.sas7bdat}, available
   from my CPH 636 home page.  The result is  2.7078221.

PROC MEANS DATA=RCLIB.FEVVALID MEAN;
VAR FEV;
RUN;


3. Calculate predicted values of the response variable for 
   subjects in the validation (or test) subset using the
   fitted model obtained by applying the REGDIAG macro to
   the training subset.  The example code below computes
   predicted values of forced expiratory volume for subjects 
   in {fevvalid.sas7bdat} using the fitted model determined
   from page 25 of {RCLIB.fevtrain27.rtf}.

DATA RCLIB.FEVVALID;
SET  RCLIB.FEVVALID;
PREDFEV = -4.33862 + 0.07972*AGE + 0.09992*HGT + 0.15754*SEX + 0.03088*SMOKE;
RUN;


4. Calculate the squared differences between the actual 
   values and predicted values of the response variable
   for the subjects in the validation (or test) subset.
   Also, calculate the squared differences between the
   actual values and the sample mean of the response
   variable for the subjects in the validation (or test)
   subset.

DATA RCLIB.FEVVALID;
SET  RCLIB.FEVVALID;
SQDIFFPRED = (FEV - PREDFEV)*(FEV - PREDFEV);
SQDIFFMEAN = (FEV - 2.7078221)*(FEV - 2.7078221);
RUN;


5. Obtain  R^2  for the validation (or test) subset via the
   formula  1 - A/B, where  A  is the sum of  SQDIFFPRED
   over all subjects in the validation (or test) subset
   and  B  is the sum of  SQDIFFMEAN  over all subjects in
   the validation (or test) subset.  In this example, we
   find that  A = 29.2145933  and  B = 130.1741638, whence
   R^2  for the validation subset in {fevvalid.sas7bdat} is
   1 - 29.2145933/130.1741638 = 0.7756.  

PROC MEANS DATA=RCLIB.FEVVALID SUM;
VAR SQDIFFPRED SQDIFFMEAN;
RUN;