Dear All!
SPSS 13 produces regression estimates of factor scores properly (imo). However the same syntax (40 items, scored from 1 to 5, PAF solution, 5 factors rotated by Promax) produces odd results when using version 14 to the same data. Factor score correlation are small negative, they all should be positive (and are in v13 solution), stardard deviations of the scores exeed 1 slightly (not in v13), so this can not be the multiple R ahcieved in estimation. Factor scores produced with same syntax using same data and versions 13 and 14 correlate abt .8 (I expected them to be identical). I give more details about syntax and data, if necessary. Something has been obviously changed. Can someone explain? Rgds Erkki -- + 358 40 5024491 <http://www.helsinki.fi/~komulain/> |
Hi!
I'll let you know that SPSS examined the case and stated: --- quote " I've confirmed that we do indeed have a problem, which is that the pattern matrix rather than the structure matrix is being used in the computation of the factor score coefficient matrix. This results in incorrect values in that matrix, the saved factor scores and the factor score covariance matrix, which are all dependent upon that computation. We apologize for the error and very much appreciate your alerting us to this problem. I'll make certain that we get it fixed as quickly as possible. " --- end quote And futher about the permission to quote: --- quote " You're welcome. Yes, but please also note the following information for the scope of the problem. This appears to affect only Releases 14.0.2 and 15.0, and to apply only to promax rotation. " David Nichols - Master Statistician SPSS Inc. Email: [hidden email] Tel: +1.312.651.3412 / Fax: +1.312.264.3412 --- end quote I thank SPSS Inc. for rapid treatment of my problem! Respectfully Erkki K. -- 040-5024491 <http://www.helsinki.fi/people/Erkki.Komulainen/> On Wed, 4 Oct 2006, Erkki Komulainen wrote: > Dear All! > > SPSS 13 produces regression estimates of factor scores properly (imo). > However the same syntax (40 items, scored from 1 to 5, PAF solution, 5 > factors rotated by Promax) produces odd results when using version 14 to the > same data. > > Factor score correlation are small negative, they all should be positive (and > are in v13 solution), stardard deviations of the scores exeed 1 slightly (not > in v13), so this can not be the multiple R ahcieved in estimation. > > Factor scores produced with same syntax using same data and versions 13 and > 14 correlate abt .8 (I expected them to be identical). > > I give more details about syntax and data, if necessary. > > Something has been obviously changed. Can someone explain? > > Rgds > Erkki > > -- > + 358 40 5024491 <http://www.helsinki.fi/~komulain/> > |
Free forum by Nabble | Edit this page |