jamovi

I have performed EFA with principal axis extraction in Jamovi and in SPSS (on correlation matrix).

If the first eigenvalue shows a small difference (11.302 in jamovi versus 11.927 in SPSS), I find huge difference in the remaining eigenvalues: on the whole 4 eigenvalues greater then 1 in Jamovi (11.30, 1.90, 1.72, 1.09) versus 10 in SPSS (11.93, 2.63, 2.58, 1.91, 1.61, 1.40, 1.31, 1.13, 1.05, 1.00).

How is it possible?
Thank you for your help.

Did you use the same rotation for both? The default rotations are different in SPSS and jamovi.

EDITED: Nevermind--I see that the issue is with [initial] eigenvalues, which won't change with rotation.

So the eigenvalues that are associated with the common factor analysis (or EFA) are the eigenvalues of the common factor solution. I think SPSS shows the eigenvalues of the covariance matrix (like in PCA in jamovi). So this could explain the difference. You can compare the SPSS eigenvalues to the PCA eigenvalues in jamovi to check whether they are the same.

Actually SPSS performs an EFA that's not a real EFA, but a PCA, and has no parallel analysis implemented. Stay away from it. One concern is about the nature of your data: ordinal or continuous? you must check the proper correlation (i.e.: polychoric, tethracoric, pearson, spearman, kendall) before decide the number of factors to retain, so on and so forth,

Ravi wrote:So the eigenvalues that are associated with the common factor analysis (or EFA) are the eigenvalues of the common factor solution. I think SPSS shows the eigenvalues of the covariance matrix (like in PCA in jamovi). So this could explain the difference. You can compare the SPSS eigenvalues to the PCA eigenvalues in jamovi to check whether they are the same.

Ravi wrote:So the eigenvalues that are associated with the common factor analysis (or EFA) are the eigenvalues of the common factor solution. I think SPSS shows the eigenvalues of the covariance matrix (like in PCA in jamovi). So this could explain the difference. You can compare the SPSS eigenvalues to the PCA eigenvalues in jamovi to check whether they are the same.

I dont think it's that because in my SPSS syntax choose METHOD = CORRELATION

mgabriel wrote:Actually SPSS performs an EFA that's not a real EFA, but a PCA, and has no parallel analysis implemented. Stay away from it. One concern is about the nature of your data: ordinal or continuous? you must check the proper correlation (i.e.: polychoric, tethracoric, pearson, spearman, kendall) before decide the number of factors to retain, so on and so forth,

Thank you for this information: the figures of SPSS EFA on my data are the same as the one from JAMOVI PCA. Do you have a reference about this problem of SPSS EFA?

Does this mean that all the articles that I have read about the validation of an instrument (in the field of psychology) with EFA SPSS are partially incorrect?