
Message/Author 


Hello Linda, could you help me with the next output of my model? THE MODEL ESTIMATION TERMINATED NORMALLY WARNING: THE LATENT VARIABLE COVARIANCE MATRIX (PSI) IS NOT POSITIVE DEFINITE. THIS COULD INDICATE A NEGATIVE VARIANCE/RESIDUAL VARIANCE FOR A LATENT VARIABLE, A CORRELATION GREATER OR EQUAL TO ONE BETWEEN TWO LATENT VARIABLES, OR A LINEAR DEPENDENCY AMONG MORE THAN TWO LATENT VARIABLES. CHECK THE TECH4 OUTPUT FOR MORE INFORMATION. PROBLEM INVOLVING VARIABLE F4. In the TECH 4 output I have found a correlation> 1 between the latent variables F1 and F6, F2 and F7, F3 and F8, F4 and F9, which should be strongly correlated according to the theoretical model but <1. In the same way I have found residual or negative variance between several latent variables F1 and F3, F1 and F8, F2 and F8, F2 and F13, F3 and F11, F7 and F13. How can I adjust my model with the theoretical without eliminating latent variables or dependent variables, since these have saturation> 0.3 with their factor? Thank you! 


Theories are often not right for the data, for instance because of the particular measurement instrument, the particular population from which the sample is drawn, or because the statistical model does not reflect the theory well enough. Try EFA to see if the factor correlations decrease due to EFA's inclusion of crossloadings. 

Jesus Garcia posted on Wednesday, November 21, 2018  12:16 pm



The data with which I am working are 80 variables to which I have already done EFA to determine the number of factors to extract, this analysis shown 15 factors each with 5 or 6 variables according to the theoretical model, however, in the CFA a the factors I have been eliminating variables that do not have a good saturation with their factor <0.4, until reaching these indicators RMSEA 0.051 CFI 0.856 TLI 0.839 SRMR 0.057. (I have to clearly adjust my model) but in the last output it shows me the following: WARNING: THE LATENT VARIABLE COVARIANCE MATRIX (PSI) IS NOT POSITIVE DEFINITE. THIS COULD INDICATE A NEGATIVE VARIANCE/RESIDUAL VARIANCE FOR A LATENT VARIABLE, A CORRELATION GREATER OR EQUAL TO ONE BETWEEN TWO LATENT VARIABLES, OR A LINEAR DEPENDENCY AMONG MORE THAN TWO LATENT VARIABLES. CHECK THE TECH4 OUTPUT FOR MORE INFORMATION. PROBLEM INVOLVING VARIABLE F11. There is no residual covariance in the sample and there are no correlations> 1, so I suppose it is a linear dependence between several latent variables. How could I solve this? Can I send you my .dat file and input/output file? Thank you very much. 


If the EFA doesn't have this PSI problem for 15 factors, it means that the CFA model fixes too many significant EFA crossloadings at zero. 

Back to top 

