Amber Fahey posted on Wednesday, November 15, 2017 - 6:05 pm
Hello, I am trying to identify the rate and trajectory of orientation in patients with moderate and severe TBI while in acute rehabilitation. Administration of test was on different days, and different number of times so I am using wide format, and commands outlined in example 6.12 of the manual. I have 20 administrations but I have to adjust coverage to 0.001. I also get a warning: WARNING: THE MODEL ESTIMATION HAS REACHED A SADDLE POINT OR A POINT WHERE THE OBSERVED AND THE EXPECTED INFORMATION MATRICES DO NOT MATCH. AN ADJUSTMENT TO THE ESTIMATION OF THE INFORMATION MATRIX HAS BEEN MADE. THE CONDITION NUMBER IS -0.183D-03. THE PROBLEM MAY ALSO BE RESOLVED BY DECREASING THE VALUE OF THE MCONVERGENCE OR LOGCRITERION OPTIONS OR BY CHANGING THE STARTING VALUES OR BY USING THE MLF ESTIMATOR.
1. Is there a recommended minimum value for coverage? 2. I have read that it is recommended to adjust sample variances to 1-10 using the DEFINE command but I am not sure which variances to look at to determine this. 3. How do I determine what to adjust the MCONVERGENCE value to, and the number of iterations, and starting values? 4. How do you adjust Logcriterion/what does the warning mean when it says to use LOGCRITERION OPTIONS? 5. What order of corrections do you suggest I do/start with?
You may want to analyze this instead in long, twolevel format as shown in UG ex 9.16.
Amber Fahey posted on Friday, November 17, 2017 - 12:44 pm
Thank you for responding so quickly! I am new to Mplus and growth modeling so I apologize for the following question, but what is the advantage to analyzing it in the long format vs the wide as I had been doing?
Long format allows you to analyze much longer longitudinal data than wide can handle. But wide gives more flexible modeling.
Amber Fahey posted on Wednesday, November 22, 2017 - 10:49 am
Thank you for your prompt response. Will the UG ex. 9.16 work without a covariate? I can't seem to figure out how to run it as an unconditional model, and not specifying anything in the Between = and %between% commands.
Send your output to Support along with your license number.
Amber Fahey posted on Sunday, February 18, 2018 - 10:10 am
Hello again, I have successfully modeled my data using the syntax in UG ex 9.16 (thank you for that suggestion). However, now I would like to run a conditional model with 13 covariates. When I run it with all the covariates, I get the following warnings:
WARNING: THE SAMPLE COVARIANCE OF THE INDEPENDENT VARIABLES IS SINGULAR. PROBLEM INVOLVING VARIABLE AGE. WARNING: THE MODEL ESTIMATION HAS REACHED A SADDLE POINT OR A POINT WHERE THE OBSERVED AND THE EXPECTED INFORMATION MATRICES DO NOT MATCH. AN ADJUSTMENT TO THE ESTIMATION OF THE INFORMATION MATRIX HAS BEEN MADE. THE CONDITION NUMBER IS -0.589D-02. THE PROBLEM MAY ALSO BE RESOLVED BY DECREASING THE VALUE OF THE MCONVERGENCE OR LOGCRITERION OPTIONS OR BY CHANGING THE STARTING VALUES OR BY USING THE MLF ESTIMATOR. THE STANDARD ERRORS OF THE MODEL PARAMETER ESTIMATES MAY NOT BE TRUSTWORTHY FOR SOME PARAMETERS DUE TO A NON-POSITIVE DEFINITE FIRST-ORDER DERIVATIVE PRODUCT MATRIX. THIS MAY BE DUE TO THE STARTING VALUES BUT MAY ALSO BE AN INDICATION OF MODEL NONIDENTIFICATION. THE CONDITION NUMBER IS 0.141D-16. PROBLEM INVOLVING PARAMETER 8.
THE MODEL ESTIMATION TERMINATED NORMALLY
I am new to Mplus and growth modeling and with several different warnings I am not sure how to proceed to check and fix errors and warnings. Your guidance is greatly appreciated!
THE PROBLEM MAY ALSO BE RESOLVED BY DECREASING THE VALUE OF THE MCONVERGENCE OR LOGCRITERION OPTIONS
As a background for this, read the FAQ on our website:
TECH8 – negative ABS changes
Amber Fahey posted on Tuesday, February 27, 2018 - 11:11 am
Thank you for the feedback. I now have a few questions regarding the change from wide to long format. If I am now running the data according to the ex. 9.16 in the UG, is this no longer considered a LATENT growth model? Do I now need to to call it a 2-level growth model? And if so, is the model no longer using a covariance structure of change? Finally, I am not sure how the random slopes affects my interpretation in comparison to say, Type = Two level (no random). Can you point me to any resources that will help me learn the difference?
See our Short Course Topic 8 video and handout, especially around slide 54.
I would call it a growth model carried out as two-level modeling. It still uses latent variables, namely the random effects random intercept and random slope.
Amber Fahey posted on Saturday, March 03, 2018 - 12:46 pm
Thank you for the suggestion! I have watched the video and studied the HO both of which I found very helpful; however, it still is not clear to me whether I have any fixed variables using the 2-level modeling with random intercept and slope and now that I understand how MPlus collapses the traditional 3 levels into 2, I am having difficulty figuring out how to count the number of parameters I have. Your suggestions are greatly appreciated!