Hi,
I have a question about conducting factor analysis on polychoric correlation. I have a correlation table in the following format in SAS:
| A | B | C | D | E |
|---|
| 1 | .1 | .2 | .5 | .4 |
| .1 | 1 | .4 | .6 | .2 |
| .2 | .4 | 1 | .2 | .4 |
| .5 | .6 | .2 | 1 | .1 |
| .4 | .2 | .4 | .1 | 1 |
A B C D E are the variables in the dataset (the actual dataset has 17 variables), with 1000 observations.
I conduct the factor analysis with the following syntax:
PROC FACTOR data =matrix1 (TYPE=CORR)
METHOD = ULS NFACTOR= 3 NOBS= 1000
ROTATE = v SCREE CORR;
RUN;
SAS complains the following:
(MESSAGE 1)
No _NAME_ variable in the TYPE=CORR data set WORK.MATRIX1. Be sure that the variables are specified in the same order as the corresponding observations appear in the dataset.
(MESSAGE 2)
The CORR matrix read from the input dataset WORK.MATRIX1 has a diagonal element equal to 0.999999 for the variable A.
The CORR matrix read from the input dataset WORK.MATRIX1 has a diagonal element equal to 0.999999 for the variable B.
The CORR matrix read from the input dataset WORK.MATRIX1 has a diagonal element equal to 0.999999 for the variable C.
The CORR matrix read from the input dataset WORK.MATRIX1 has a diagonal element equal to 0.999999 for the variable D.
The CORR matrix read from the input dataset WORK.MATRIX1 has a diagonal element equal to 0.999999 for the variable E.
So, SAS does not allow any 1's in a correlation matrix? When A correlates with A, for example, the correlation will naturally be 1. Why is SAS complaining?
Any help will be useful.
Thanks!
Chester
