In my EFA, Bartlett's Test of Sphericity is 0.000, means a significant degree of correlation exists and the data is suitable for factor analysis. However, the Determinant = 1.94E-014, indicating an identity matrix and making the data unfit for factor analysis. How indicates a high level of multicollinearity of data. Am I correct in the interpretation? If so, these are conflicting interpretations of same data. How do I reconcile this? Will deeply value your insights.
A determinant that small suggests that some of the variables included are very strongly correlated. One option is to look at the pairwise correlation coefficients between the variables you want to include. If you find any high correlations between pairs of variables you could consider removing one of the pair. Andy Field put a handout on EFA on his webpages which you might find useful as it deals with singularity / multicollinearity issues; you can find it here: http://users.sussex.ac.uk/~andyf/factor.pdf
Many thanks Chris for your inputs. I see the two (Determinant and Bartlett's Test) provides different and non-conflicting interpretations. I thank you for your insights and will check for high correlation variables for removing them.
Regards,
Vikas.
- Badges
- Science topic
- Similar topics
- Mathematics
- Algebra
- Matrix