Exploratory factorial analysis - Determinant of the correlation matrix

[Laura Galarza]

Hi guys!
I'm trying to run an EFA after adapting a questionnaire from another language. The correlation matrix looks good, with all correlations being more than 0,3 and less than 1, but the determinant of the correlation matrix is 0. What can I do now? Can I still do the EFA? KMO and Barlett values are good too.
Thank so much for your help.
Laura

[Erik Ruzek]

What are you correlating? Item values? Are the responses Likert-scaled? Are you looking at Pearson correlation coefficients? Tetrachoric (if dichotomous) or polychoric (if polytomous - search polychoric)?

[Laura Galarza]

Thanks for you answer.
I'm correlating different items from a questionnaire. The answers for this items are Likert-scaled from 1 to 5.
I run first pwcorr, and all the correlations between variables(items) are between 0,3 and 1.
Then I run factortest to obtain the KMO, Barlett and determinant of the correlation matrix. KMO and Barlett values are good, but the determinant is 0.
Honestly, this is the first time I do this in real life (not in courses) and not sure what to do with this determinant (and I can't find the answer in books).
Thanks

[Erik Ruzek]

If you have two or more variables that are very highly correlated (perhaps > 0.96), you could end up with a determinant value of 0. You mention having correlations "between 0,3 and 1." How close to 1 are we talking? I would try removing one of the variables in a pair w/ a correlation > 0.96 and see if this changes the determinant value from factortest. Another way of detecting this problem is by running a PCA on the variables; you will likely see the last eigenvalue gets very close to 0 if not becomes 0.

[Laura Galarza]

Thanks so much EriK. I'll try your solution and see if it works.