topic Re: Factor Analysis on pre-defined correlation matrix in New SAS User

Factor Analysis on pre-defined correlation matrix

lmpsumagpang — Mon, 16 Mar 2020 10:37:30 GMT

Is it possible to run a Factor Analysis on a pre-defined correlation matrix? From what I know, if you run PROC FACTOR, the first thing SAS does is create a correlation matrix but since the dataset I have is already the correlation matrix, wouldn't that mess up the whole process?

My code:

/* check if correlation coefficients are the same as the original coefficients in the dataset */
proc corr data=dataset outs=datacorr; run;

/* Factor Analysis */
ods graphics on;
proc factor data=datacorr r=v n=3 plots=loadings;
run;
ods graphics off;

Re: Factor Analysis on pre-defined correlation matrix

PaigeMiller — Mon, 16 Mar 2020 10:52:13 GMT

This is a great example of why you should look at the documentation.

A correlation data set (with option TYPE=CORR) is valid input to PROC FACTOR and PROC PRINCOMP. Since your output is from PROC CORR, there should be no issue using it in PROC FACTOR or PROC PRINCOMP. Details are here: https://documentation.sas.com/?cdcId=pgmsascdc&cdcVersion=9.4_3.4&docsetId=statug&docsetTarget=statug_factor_details01.htm&locale=en

Re: Factor Analysis on pre-defined correlation matrix

lmpsumagpang — Mon, 16 Mar 2020 10:57:30 GMT

Thanks but I was wondering why I have to compute for a new correlation matrix when my initial dataset is already a correlation matrix. Would the analysis significantly change?

Is there a way where I can import the initial dataset and dictate it's TYPE (CORR) so SAS won't have to compute a new correlation matrix from it?

Re: Factor Analysis on pre-defined correlation matrix

PaigeMiller — Mon, 16 Mar 2020 11:01:16 GMT

@lmpsumagpang wrote:

Thanks but I was wondering why I have to compute for a new correlation matrix when my initial dataset is already a correlation matrix. Would the analysis significantly change?

As far as I know, you don't have to compute a new correlation data set if you have PROC CORR output. That's my understanding, and the link I gave seems to say that as well. This is what it says:

The data set created by the CORR procedure is automatically given the TYPE=CORR data set option, so you do not have to specify TYPE=CORR.

Re: Factor Analysis on pre-defined correlation matrix

lmpsumagpang — Mon, 16 Mar 2020 11:06:15 GMT

Please disregard the proc corr code that I shared in my inquiry.

The original data set that I've imported (from an excel file) is already a correlation matrix. Making a proc factor function will result to SAS making a new correlation matrix out of it which might affect my analysis. Is there a way where I can import the excel file and set it as TYPE=CORR?

Thanks for your immediate response and I appreciate all the help!

Re: Factor Analysis on pre-defined correlation matrix

PaigeMiller — Mon, 16 Mar 2020 11:19:36 GMT

Rough outline of code

PROC IMPORT datafile = 'whatever.xlsx' out=a dbms=excel;

run;

data corr(type=corr);

set a;

run;

proc factor data=corr;

...

run;

DISCLAIMER: this may not work on correlation data sets stored in Excel, as there's no standard that I know of for correlation matrices in Excel being formatted/laid out the same as a PROC CORR output data set. You can't know until you try it.

Re: Factor Analysis on pre-defined correlation matrix

lmpsumagpang — Mon, 16 Mar 2020 11:44:28 GMT

Thank you! It worked fine.

The excel file had the same format and data type as the format using the proc corr statement