Today I was conducting a method factor model to test for common method bias. I ran the model in Mplus and then in AMOS to check my results. Both models produced the exact same parameter estimates, df, chi-square, etc. However, the two programs report a different number of free parameters.
To isolate the issue, I broke my model down to a simple CFA with 1 latent factor and 8 observed variables. In AMOS and in my head, that results in 16 free / estimated parameters. Hence 7 measurement relationships (8th is fixed @1) + 8 residual variances + 1 variance = 16.
However, Mplus says there are 24 free parameters. The difference between Mplus and AMOS is 24-16 = 8 (i.e. the number of observed variables I am working with).
Does anyone know why Mplus is getting 24 free parameters? What am I missing here? It's been a long day. Is Mplus counting the intercepts?
Thanks so much!
ANALYSIS: TYPE IS GENERAL; ESTIMATOR IS ML; ITERATIONS = 1000; CONVERGENCE = 0.00005;