 Cat factor indicators - item-total co...    Message/Author  Elina Dale posted on Sunday, April 21, 2013 - 8:10 pm
Dear Dr. Muthen,

I am trying to obtain item-total correlations as described in Marcoulides & Raykov (2011),p. 109. But I am not sure where in the output I should see the correlation matrix. I put in the following commands:
CATEGORICAL = intr1 intr3 intr7 intrj4 ;
USEVARIABLE = intr1 intr3 intr7 intrj4 f1t ;
Define: f1t = intr1 + intr7 + intrj4 ; !OBTAIN ITEM-TOTAL CORR FOR INTR3
Analysis: TYPE = BASIC ;

In the output under RESULTS FOR BASIC ANALYSIS, I see the following matrix:
CORRELATION MATRIX (WITH VARIANCES ON THE DIAGONAL)

This matrix contains all of the items, including the one I have newly defined. However, the diagonal cells are empty, except for the newly defined var f1t. Is this supposed to be a CORRELATION matrix or a Covariance matrix? Did I specify commands correctly in the input file (above) to get the correlation matrix? If it is a correlation matrix like I was hoping, then why does my newly defined item have a value 1.2 in the diagonal cell? All the rest of the items have empty cells in the diagonal.

As I understand, since I have categorical variables, I get polychoric correlation matrix. Or since my total score f1t is a sum, it's continuous and that is why I am getting that strange number on the diagonal?

Thank you!!!  Bengt O. Muthen posted on Monday, April 22, 2013 - 8:35 am
I think you refer to the 2011 Raykov-Marcoulides book Introduction to Psychometric Theory.

As the output says, this is a matrix of correlations with variances on the diagonal. Categorical variables do not have variance parameters estimated so the diagonal elements are empty. You specify the analysis correctly and you are correct that the sum gives a variance of 1.2, while the off-diagonal elements are polychorics.    Topics | Tree View | Search | Help/Instructions | Program Credits Administration