Abby Gray posted on Thursday, November 21, 2013 - 11:35 am
I am specifying a CFA model for MTMM data (N=196). I have tried several approaches (CTCM, CTCU, and CT-C(M-1)). The basic model (which does not address common-method variance) fits very well, but these MTMM models fail to converge or generate error messages about identification problems. Do you have any suggestions about how I might improve my chances of successfully running an MTMM model?
Thank you very much for any suggestions. Here is the script and error message from one of my many failed attempts (this is the CT-C(M-1) attempt):
USEVARIABLES are SC1 SC2 TL5 RR1 RR2 RR3 FG1 FG2 FG3 FG6 FG8 FG9 FG10; MISSING are all (-999);
ANALYSIS: Estimator is MLR; MODEL: F1 by SC1* SC2 TL5; F2 by RR1* RR2 RR3 ; F3 by FG1* FG2 FG3 ; F4 by FG6* FG8 FG9 FG10 ;
! method factors below. F5 by SC1* SC2; F6 by RR1* RR2 RR3; F7 by Tl5*;
Error message: THE MODEL ESTIMATION TERMINATED NORMALLY THE STANDARD ERRORS OF THE MODEL PARAMETER ESTIMATES COULD NOT BE COMPUTED. THE MODEL MAY NOT BE IDENTIFIED. CHECK YOUR MODEL. PROBLEM INVOLVING PARAMETER 22. THE CONDITION NUMBER IS -0.173D-01. THE ROBUST CHI-SQUARE COULD NOT BE COMPUTED.
Shirley posted on Monday, September 05, 2016 - 10:09 pm
Dear Dr. Muthen, I am investigating the factorial structure of a survey. Some of the items are negatively worded, and I am considering to model this method effect in CFA framework.
As a start, I have tried to fit a MTMM model in CFA framework with 5 traits and 2 method factors (positively vs. negatively worded). The model fit well, but part of the solution is difficult to explain.
I recently read from Brown (2006, p.218)* that there must be at least 3 traits and 3 methods for such a MTMM model to be identified. I am wondering if this means I shouldn't trust the output of the 5-trait-2-method CFA model, although the model did converge successfully.
Would appreciate your advice. Thanks.
Brown, T.A. (2006). Confirmatory factor analysis for applied research. New York: Guilford Press.