Anonymous posted on Tuesday, July 12, 2005 - 2:46 pm
I am working on a model that has dichotomous items defining a factor, which runs in Mplus just fine. Then when I try to add a gender variable with a relationship to the factor (only), I get this error message...
THE MODEL ESTIMATION DID NOT TERMINATE NORMALLY DUE TO AN ILL-CONDITIONED FISHER INFORMATION MATRIX. CHANGE YOUR MODEL AND/OR STARTING VALUES.
THE MODEL ESTIMATION DID NOT TERMINATE NORMALLY DUE TO A NON-POSITIVE DEFINITE FISHER INFORMATION MATRIX. THIS MAY BE DUE TO THE STARTING VALUES BUT MAY ALSO BE AN INDICATION OF MODEL NONIDENTIFICATION. THE CONDITION NUMBER IS -0.981D-12.
THE STANDARD ERRORS OF THE MODEL PARAMETER ESTIMATES COULD NOT BE COMPUTED. THIS IS OFTEN DUE TO THE STARTING VALUES BUT MAY ALSO BE AN INDICATION OF MODEL NONIDENTIFICATION. CHANGE YOUR MODEL AND/OR STARTING VALUES. PROBLEM INVOLVING PARAMETER 1.
I have tried using the starting values from previous item level runs without gender (which worked fine), and I am getting the same error. Since I am still learning about Mplus and item level factor analysis, I am not exactly sure what is wrong with the model itself. Is there a certain type of error that this message points to (i.e missing a variance or factor loading constraint)? I am just not sure where the nonidentification would be.
This is the type of problem that needs to be sent to firstname.lastname@example.org along with the input, data, output, and license number. Diagnosinig data specific problems is difficult without this information.
Maggie Chun posted on Friday, March 07, 2008 - 11:16 am
I am a new user for Mplus.
I tried to run multiple group CFA for a scale with 7 item on one factor, and got the IRT 2P report, but I do not know what are the index for the four columes number under "Item discriminations" and another four under "item difficulties". To compare the item function between two groups, which index I should use?
I have learned a lot from your webpage. Thank you very much!!
The four columns have the same headings as the rest of the results. You can compare the item function using both the difficulty and discrimination parameters and the icc plot which summarizes them.
Maggie Chun posted on Tuesday, March 11, 2008 - 6:50 am
Thank you very much, Linda.
Have a nice week!
nanda mooij posted on Wednesday, June 23, 2010 - 6:03 am
Dear Dr. Muthen,
I red this question above about item difficulties en item discriminations. Apparantly, she gets these in her output. While I was thinking that i have to calculate the difficulties en discriminations myself with the formulas of the webnote4. I am running this model: VARIABLE: NAMES ARE v1-v144; USEVARIABLES ARE v1-v16; CATEGORICAL ARE v1-v16; MISSING ARE .; ANALYSIS: estimator=wlsm; MODEL: f1 BY v1-v16; f2 BY v17-v32; f3 BY v33-v48; h1 BY f1-f3; The output says: parametrization=delta. I'm using Mplus 4. So is this version not able to give the difficulties and discriminations, or do I have to fit the model differently?
Thanks for the answer. I'm translating the parameters myself now and I have a question about the formula for the a-parameter for delta parameterization. In the formula, there is a delta sign, and I'm wondering where in the ouput I can find these number(s)? Thanks in regard, Nanda
Hi Can you please let me know what is the problem with my model when all the observed variables are significantly loading but the posterior predictive distribution value is constantly getting significant: The model fit for CFA with 4 latent and 20 dependent variables (all categorical, mostly binary) is as follows:
MODEL FIT INFORMATION
Number of Free Parameters 38
Bayesian Posterior Predictive Checking using Chi-Square
95% Confidence Interval for the Difference Between the Observed and the Replicated Chi-Square Values