Extracting fit statistics from MI output PreviousNext
Mplus Discussion > Missing Data Modeling >
Message/Author
 Shin, Tacksoo posted on Wednesday, November 30, 2011 - 5:22 am
Dear Linda and Bengt,

I am doing missing data study with Bayesian based analysis (Multiple imputation). For this simulation study, MplusAutomation of R program is also imposed. I successfully created and run models. Problem is that the model fit statistics (e.g., Posterior fit statistics and DIC) cannot read within R console. When reading Model output (readModels in R), there were only parameter estimate information. Also, when running summaryStats in R, I failed to obtain any model fit information.
But, I found all fit results in the separated Mplus outputs.

Do you have any idea about it? Am I wrong? in what respect?

Second question is "PPP can be interpreted same as p-value of chi-square in ML analysis?" Can they be comparable?

Thank you
 Linda K. Muthen posted on Thursday, December 01, 2011 - 2:53 pm
I checked with the author of MplusAutomation and he had not added the fit statistics. He is doing this now and it will be in the next version of MplusAutomation which will be available next week. You can contact the author directly if you have further questions.

Yes, the p-value of PPP can be interpreted the same as the chi-square p-value. For further information, see on the website under Papers:

Muthén, B. & Asparouhov, T. (2011). Bayesian SEM: A more flexible representation of substantive theory. Forthcoming in Psychological Methods.
 Shin, Tacksoo posted on Friday, December 02, 2011 - 2:48 pm
Dear Linda,

Truly appreciate your support. I always learn a lot for this discussion board. Thank you!

Warmly,
Tacksoo
 Shin, Tacksoo posted on Sunday, June 30, 2013 - 8:23 pm
Dear Linda,

When extracting model fit information through MplusAutomation by R, "showSummaryTable" easily organizes and prints all outputs. Is there any similar option when extracting model parameter estimates?

Thank you.
 Linda K. Muthen posted on Monday, July 01, 2013 - 7:49 am
You should contact the MplusAutomation author regarding this question.
 Shin, Tacksoo posted on Thursday, July 11, 2013 - 3:00 pm
Dear Linda,

I found that the recent version (7.1) of Mplus provides BIC in Bayesian output. So, I tried to extract BIC values through MplusAutomation. Unfortunately, I got an error message.

R command: showSummaryTable(summaryStats, keepCols=c("Title", "DIC", "BIC", "PostPred_PValue"), sortBy="BIC")

Error Message is followed,
Error in subsetModelList(modelList, keepCols, dropCols, sortBy) : sortBy field: BIC is not present in the summary data.frame.

Should I update and download a different version of MplusAutomation? Or this fit statistic should be added in program?

Always thanks for your help...
 Michael Hallquist posted on Friday, July 12, 2013 - 7:19 am
Hello Tacksoo,

It is hard to diagnose the problem you described without seeing the Mplus output file or the structure of the R data read in by MplusAutomation. But if you're extracting summary statistics from a multiple imputation analysis, then most of them are reported as means and SDs. So, I would do this first:

str(summaryStats)

to see what's in the data.frame. I would guess you'll see something like:

Classes 'mplus.summaries' and 'data.frame': 1 obs. of 38 variables:
$ Mplus.version : chr "6"
$ AnalysisType : chr "complex"
$ DataType : chr "IMPUTATION"
$ Estimator : chr "MLR"
$ Parameters : num 15
$ LL_Mean : num -12270
$ LL_SD : num 26.5
$ LL_NumComputations : num 5
$ AIC_Mean : num 24570
$ AIC_SD : num 53.1
$ AIC_NumComputations : num 5
$ BIC_Mean : num 24665
$ BIC_SD : num 53.1

If that's the case, you should be passing things like "DIC_Mean" and "BIC_Mean" to showSummaryTable.

If this doesn't solve the problem, please email me the output file and all of the relevant R code.

Hope that helps,
Michael
Back to top
Add Your Message Here
Post:
Username: Posting Information:
This is a private posting area. Only registered users and moderators may post messages here.
Password:
Options: Enable HTML code in message
Automatically activate URLs in message
Action: