Hirotaka Jo posted on Saturday, April 07, 2012 - 9:45 am
Dear Professor Muthen.
I have two questions concerning EFA.
First question As my data is dichotomous, I calculate Tetrachoric correlation coefficient. The output shows some messages below WARNING: THE BIVARIATE TABLE OF V5 AND V1 HAS AN EMPTY CELL. The correlation coefficient of V5 and V1 of Mplus output is 0.979. The result of same variables in R (freeware) was 0.999. Why these values are different? Does Mplus replace tetrachoric correlation to other correlation?
Second question I did factor analysis using SPSS. Calculation stopped because of linear combination. But EFA of Mplus completed analysis. Why Mplus can calculate? In addition, I donft know that it is possible to use Mplus result for my thesis because of below message THE INPUT SAMPLE CORRELATION MATRIX IS NOT POSITIVE DEFINITE.
A zero cell in a bivariate table implies a correlation of one. There are different ways to deal with zero cells resulting in correlations being estimated at different values. If you have a zero cell, you should use only one of the variables in the analysis if the model is being fitted to tetrachoric correlations. Alternatively, you can use maximum likelihood with categorical outcomes which fits the model to the raw data.
This message can be ignored. You would get a different message if some of your variables are from a linear combination.