Chi-square in MPlus vs. Amos
Message/Author
 Mary Beth Oliver posted on Tuesday, March 02, 2010 - 6:38 am
Hello Everyone,

I'm new to MPlus (had been using Amos), so apologies upfront if this is a misplaced question.

In trying to learn MPlus, I'm running some very basic analyses in MPlus and Amos so I can compare the results. I'm just running a single factor CFA.

I'm stumped because the chi-square for the model differs for the two outputs, but all of the fit statistics, DF, etc. are identical (or very, very close to it). As a consequence of this, I believe, things like the modification indices are different. I'm using ML estimation.

There's not a HUGE difference in the chi-squares, but some: For example, with a sample size of 399:

In Amos: 89.042, DF=9
In MPlus: 89.266, DF=9

The fit for the baseline model is slightly different, too.

I'll paste my Mplus code below.

Any thoughts would be most appreciated!

With kind regards,

Mary

***********************

DATA:
FILE IS "religion just world.dat";

VARIABLE:
NAMES ARE subject rel1 rel2 jw1 jw2 jw3 jw4 jw5 jw6;
USEVARIABLES ARE jw1 jw2 jw3 jw4 jw5 jw6;

ANALYSIS:
ESTIMATOR IS ML;

model:

jwf by jw1-jw6;

OUTPUT: SAMPSTAT MODINDICES (ALL) STANDARDIZED CINTERVAL FSCOEFFICIENT TECH3;
 Linda K. Muthen posted on Tuesday, March 02, 2010 - 6:44 am
Mplus uses n for chi-square. Amos uses n-1.

The baseline model in Mplus is the means/thresholds and variances of all observed variables and the covariances among the observed exogenous variables. In Amos, I believe it is means/thresholds and variances of all observed variables with the covariances among the observed exogenous variables fixed at zero.
 Mary Beth Oliver posted on Tuesday, March 02, 2010 - 11:47 am
Thank you so much for your quick and very helpful response. Problem solved! Fantastic. Thank you!
 Rachel Kane posted on Tuesday, September 17, 2013 - 12:08 pm
Similar to Mary Beth's question above, I am new to MPlus and have run a CFA in AMOS and in MPlus.

ANALYSIS: TYPE = general;
MODEL:
NegAff by NA1@1 NA2 NA3 NA4;
Sat by sat1@1 sat2 sat3;
Frustr by frust1@1 frust2 frust3;
InterJus by IJ1@1 IJ2 IJ3 IJ4;
OUTPUT: stdyx;

The Fit statistics and degrees of freedom are VERY different.

AMOS:
X2= 128.31, df = 71
CFI = .978
TLI = .971
RMSEA = .056

MPLUS:
X2= 8470.33, df = 133
CFI = .262
TLI = .151
RMSEA = .456

I am getting this error message, which may have something to do with the results? I'm not quite sure.

"*** WARNING in MODEL command
Variable is uncorrelated with all other variables: NA
*** WARNING in MODEL command
Variable is uncorrelated with all other variables: JOBSAT
*** WARNING in MODEL command
Variable is uncorrelated with all other variables: FRUST
*** WARNING in MODEL command
Variable is uncorrelated with all other variables: INTJUST
*** WARNING in MODEL command
At least one variable is uncorrelated with all other variables in the model.
Check that this is what is intended."

Any guidance would be appreciated!!

Thanks,

Rachel
 Linda K. Muthen posted on Tuesday, September 17, 2013 - 12:23 pm
It sounds like in Mplus you have variables on the USEVARIABLES list that are not used in the mMODEL command. The USEVARIABLES list should have only analysis variables on it.
 Rachel Kane posted on Tuesday, September 17, 2013 - 5:12 pm
Thanks for the quick response. That explains it!!