Message/Author |
|
|
Dear sir, I was trying to do a simulation study on reliability estimation. I generate a bi-factor ESEM model and used the same model to fit the data, but the fix indexes are very poor (CFI TLI <0.5,>0.3). I ran several models and it all get the same result. Could you please check the following code for example? Data generation code: MODEL POPULATION: [x1-x10@0]; x1-x10@1; g@1; f1-f2@1; g WITH f1-f2@0; f1 WITH f2@0; g by x1-x10@0.4; f1 by x1-x5@0.3; f2 by x6-x10@0.3; f1 by x6-x10@0.001; f2 by x1-x5@0.001; Data analysis code: (output) MODEL: g WITH f1-f2@0; f1 WITH f2@0; g@1; f1-f2@1; g by x1-x10@0.4; f1 by x1-x5@0.3; f2 by x6-x10@0.3; f1 by x6-x10@0.001; f2 by x1-x5@0.001; VAR(VARX); RMSEA 0.494; CFI 0.039; TLI 0.039 |
|
|
In your Model command you should not fix all parameters - for parameters you want to estimate, you want to change @ to *. |
|
Gabriella Fu posted on Thursday, November 08, 2018 - 7:03 pm
|
|
|
Thank you for your reply. I changed the analysis code to the following (the label is for calculation): ANALYSIS: ROTATION=BI-GEOMIN(ORTHOGONAL); ITERATIONS=10000; MODEL: g by x1-x10*(l1-l10); f1-f2 by x1-x10(*1); f1 by x1-x10*(la1-la10); f2 by x1-x10*(lb1-lb10); x1-x10(v1-v10); g@1; g WITH f1-f2@0; I repulicate the model 500 times and not once did the model converge to a proper solution. Is there something wrong with the code? |
|
Gabriella Fu posted on Thursday, November 08, 2018 - 7:08 pm
|
|
|
Mostly the error reads: NO CONVERGENCE. NUMBER OF ITERATIONS EXCEEDED. WARNING: THE RESIDUAL COVARIANCE MATRIX (THETA) IS NOT POSITIVE DEFINITE. THIS COULD INDICATE A NEGATIVE VARIANCE/RESIDUAL VARIANCE FOR AN OBSERVED VARIABLE, A CORRELATION GREATER OR EQUAL TO ONE BETWEEN TWO OBSERVED VARIABLES, OR A LINEAR DEPENDENCY AMONG MORE THAN TWO OBSERVED VARIABLES. CHECK THE RESULTS SECTION FOR MORE INFORMATION. PROBLEM INVOLVING VARIABLE X3. THE CHI-SQUARE STATISTIC IS NEGATIVE. THE LOGLIKELIHOOD VALUES MAY NOT BE RELIABLE. |
|
|
Send your output to Support along with your license number. |
|
Back to top |