I'm interesting in undertaking Factor Analysis on ordinal data using SAS EG. One option that seems to have a very significant effect on the results reads -
'Correct the resulting matrix for the mean'
When is it appropriate to do this?
If I do correct the resulting matrix the correlations that come out of the Factor Analysis are of similar magnitude to a Spearman Correlation on the same data and the Factor Analysis yields what seems to be sensible and expected results.
If I do not correct the resulting matrix the correlations that come out of the Factor Analysis are nearly all over 0.90 and the Factor Analysis yields only one factor common to all variables that explains all the variance...