|
At 13:15 26/01/99 +0000, Peter Crawford wrote:
>Isn't this what proc score is for
Peter, it is indeed - and all Maria needs to do to use it is to add a couple
of variables (_TYPE_ and _NAME_ - they *must* have these names) to her
'coefficients' data set. In Maria's case, _TYPE_ will always be SCORE and
_NAME_ will always be the name she wants for the variable representing the
predicted value. So, in terms of the simple example she gave:
>So lib.estim has the following structure,
> var1 var2 var3 var4
> 0.1 0.2 0.3 0.4
>where varI is the name of the predictor variable and 0.1 etc is the
>corresponding regression coefficient estimate
>And lib.data :
>>var1 var2 var3 var4
>1 3 4 5
>2 3 6 7
>8 9 10 11
>where varI is again the predictor name and the variable values are the
>predictor values.
>The simplest and extremely time consuming way is to calculate predicted
>value by typing in the coefficients in the equation by hand.
>The simplest and extremely time consuming way is to calculate predicted
>value by typing in the coefficients in the equation by hand.
>data lib.data;
>pred=0.1*var1+0.2*var2+0.3*var3+0.4*var4;
>run;
>The problem is that I will have to do it dozens of times and I have over
>40 predictors.
data lib.estim ;
set lib.estim ;
_TYPE_ = 'SCORE' ;
_NAME_ = 'pred' ;
run ;
proc score data = lib.data score = lib.estim out = whatever ;
var var1 - var4 ;
run ;
If only the predicted value is wanted in the output dataset ('whatever'),
then one can obviously add a dataset option (keep = whatever) after its name.
Regards,
John
----------------------------------------------------------------
Dr John Whittington, Voice: +44 (0) 1296 730225
Mediscience Services Fax: +44 (0) 1296 738893
Twyford Manor, Twyford, E-mail: medisci@powernet.com
Buckingham MK18 4EL, UK mediscience@compuserve.com
----------------------------------------------------------------
| 
