# How can I apply models to predict the output of a test dataset using SAS?

I'm trying to use SAS from more of a machine learning perspective than a pure stats perspective. I want to perform resampling on a dataset to measure the predictive accuracy of a LDA and Logistic Regression. However, when I perform proc logistic or proc discrim all SAS seems to output is a bunch of statistical measures. I want to use the models to predict the output of my test dataset. Is there any way to do this in SAS?

Section 4.4 in the SAS course notes on Multivariate Statistical Methods: Practical Research Applications there are examples for empirical validation and scoring.

You'd want to consider using the testlist or testout options for example using proc discrim

proc discrim data = old-data testdata = new-data testlist;
class variabls;
priors priors;
var variables;
run;

or

proc discrim data = old-data testdata = new-data testout = scored-data;
class variable;
priors priors;
var variables;
run;

You should look into the OUTPUT statements on both those PROCs, and also look at the SCORE statement, for using a model on new data. More on SCORE from SAS: http://support.sas.com/rnd/app/da/new/801ce/stat/chap4/sect11.htm