Factor Scores with Factor Analysis using tetrachoric correlations

[emaneman]

Hello all,

I am conducting a factor analysis on a matrix of tetrachoric correlations (since my vars are binary).

I create the matrix using PROC CORR, and then use that as input dataset for PROC FACTOR.

I would like, however, compute factor scores for the two factors that I extract, and have them added to the original dataset (DATA).

Proc factor does not do that when the input is not raw data but a matrix of correlations. I can of course extract coefficients, and then do it "manually". 

Is there a way to do this automatically?

Syntax below.

Thank you all in advance.

Eman

proc corr data = DATA polychoric out=TETCORR;
var x1-x10;

proc factor data=TETCORR (type=corr) nfactors=2 method=prin rotate=oblimin ;
var x1-x10;

[sbxkoenk]

Use ODS OUTPUT (ODS = Output Delivery System) ...

• ODS Output FinalCommunWgt = work.FinalCommunWgt ;
  (or maybe another table ... the one containing what you want)

ODS OUTPUT: Store any statistic created by any SAS procedure

By Rick Wicklin on The DO Loop January 9, 2017
https://blogs.sas.com/content/iml/2017/01/09/ods-output-any-statistic.html

Here's the list of ODS table names (tables that PROC FACTOR can produce via ODS OUTPUT statement ... depending on the setting and options)

SAS/STAT User's Guide

The FACTOR Procedure
https://go.documentation.sas.com/doc/en/pgmsascdc/9.4_3.5/statug/statug_factor_details15.htm

What is polychoric correlation? 
By Rick Wicklin on The DO Loop December 11, 2023
https://blogs.sas.com/content/iml/2023/12/11/polychoric-correlation.html

BR, Koen

[emaneman]

hello Koen,

thanks for your feedback. What I need cannot be directly outputted by the PROC FACTOR in my case, since I use a matrix of correlation and not raw data. 

I think, however, that i have found the solution. The last file produced(FATSCORE2) in the following sequence includes the factor scores for factor 1 and factor 2.

proc corr data = DATA polychoric out=TETCORR noprint;
var x1-x10;

proc factor data=TETCORR (type=corr) nfactors=2 method=prin rotate=oblimin score outstat=P2 ;
var x1-x10;

PROC STANDARD DATA=DATA MEAN=0 STD=1 OUT=DATAst;
VAR
var x1-x10;

proc score data=DATAst score=P2 out=FATSCORE2;
var x1-x10;
run;

[emaneman · Accepted solution]

CORRECTION

In the previous post the syntax for PROC CORR contained a mistake.

The option to output the correlation matrix should be:

OUTPLC=TETCORR (instead of out=TETCORR).