Technical Report #8:
Scoring An Oblique Principal Component
Scoring An Oblique Principal Component
Bruce Ratner, Ph.D.
Oblique Principal Components Analysis (OPCA) is a powerful exploratory data analysis technique, which can be used to uncover unexpected relationships among many variables. This report provides a SAS-code program for performing OPCA, and for scoring oblique principal components on an external dataset.
I provide an illustration of typical OPCA output along with the SAS-code program which produced it. The program should be a welcomed entry in the toolkit of data analysts who frequently work with BIG data.
********** SAS-code Program **********
Data set IN is found in Technical Report #6.
PROC VARCLUS data= IN MAXC=4 simple outstat=coef;
var GENDER_F GENDER_M MARITAL_M MARITAL_S;
run;
/* Seeking the Three Cluster Solution */
data Coef3;
set Coef;
if _ncl_ = . or _ncl_ = 3;
drop _ncl_;
run;
PROC SCORE data=IN score=Coef3 out=scored;
var GENDER_F GENDER_M MARITAL_M MARITAL_S;
run;
/*** Assigning the Individual to the Classified Cluster-Segment ***/
data scored_classified;
set scored ;
temp=max(clus1, clus2, clus3);
if clus1 = temp then predictd = clus1;
else if clus2 = temp then predictd = clus2;
else if clus3 = temp then predictd = clus3;
run;
data scored_classified (drop=temp);
set scored_classified;
temp=max(clus1, clus2, clus3);
if clus1 = temp then segment = 'clus1';
else if clus2 = temp then segment = 'clus2';
else if clus3 = temp then segment = 'clus3';
run;
1 800 DM STAT-1, or e-mail at br@dmstat1.com.
br@dmstat1.com574 Flanders Drive / North Woodmere, NY 11581 / U S A
Voice 1-516-791-3544 / Fax 1-516-791-5075
Toll Free 1 800 DM STAT-1