Monte Carlo Simulations PreviousNext
Mplus Discussion > Latent Variable Mixture Modeling >
Message/Author
 Renee McDonald posted on Tuesday, June 07, 2011 - 2:51 pm
Hello,
I am doing a simulation study. I simulated the data (1000 data files) external to Mplus. I would like to run mixture modeling on each of the files and then save the results of each of the simulated data files in one summary file with individual statistics (no averages). Furthermore, I only want to save the BIC. Is there a wait to specify such. I believe the Runall will care for saving the information in one file. However, is there anyway to specify what to save? Thank you.
 Linda K. Muthen posted on Tuesday, June 07, 2011 - 6:02 pm
I don't know of a way to do this in Mplus. See the website under How-To in the left column. You may be able to do it via R.
 Renee McDonald posted on Thursday, June 09, 2011 - 11:54 am
Thank you. Is there a Runall compatible with Windows 7? One last question. I cannot get the savedata option to work. Can you please correct coding? Thanks.

Title: Model One

Data: File=DC9Test.txt;
Type=Montecarlo;

Variable: Names are X1-X4;
Classes=c(2:4);


Analysis: Type=mixture;
Starts 100 15;

Model: %Overall%
F1 by X1 - X4;

Savedata: file is "results.txt";
save estimates;
 Stata posted on Thursday, June 09, 2011 - 12:13 pm
Can I use ordinal variable instead of dichotomous variable for latent class factor analysis? Thank you.
 Linda K. Muthen posted on Thursday, June 09, 2011 - 1:33 pm
Renee:

We no longer support RUNALL. Please see the user's guide for the correct specification of the ESTIMATES option.
 Linda K. Muthen posted on Thursday, June 09, 2011 - 1:34 pm
Stata:

Yes.
 Renee McDonald posted on Thursday, June 09, 2011 - 3:07 pm
None of the save options seem to work with the type=montecarlo. Is there anyway possible to save the results from the individual runs (in one file preferably or several files) instead of just the averages that are given by default. Thank you.
 Linda K. Muthen posted on Thursday, June 09, 2011 - 5:11 pm
Please send your output and license number to support@statmodel.com.
 Mary M Mitchell posted on Friday, April 13, 2018 - 10:15 am
Dear Drs. Muthen,

I am running a Monte Carlo Simulation with a LTA model to test pathways between covariates and transitions between classes. The link function is logit, so that means I am testing a log odds, right? I can then exponentiate that value to present an odds ratio, I assume.

Thanks,

Mary
 Bengt O. Muthen posted on Friday, April 13, 2018 - 4:15 pm
Yes.
 Mary M Mitchell posted on Sunday, April 15, 2018 - 12:16 pm
Thank you!
 Mary M Mitchell posted on Sunday, April 15, 2018 - 5:31 pm
I have another question on this Monte Carlo simulation that I'm running. I have 3 classes and 3 points in time with a binary covariate. I want to compare power estimates between a covariate that has a 50/50 split compared to one with a 90/10 split. There is a line of code: [cg#1*.5]. Does this signify that the binary covariate has a mean of .5, therefore at 50%? Or, do I alter the line of code that has [g$1@15] and change the threshold? Then, the problem is that I don't know what thresholds match with a 90/10 split.

Thanks for your help!

Mary
 Bengt O. Muthen posted on Tuesday, April 17, 2018 - 3:58 pm
See the description of the Cutpoints option in UG ex 12.1.
 Mary M Mitchell posted on Wednesday, April 18, 2018 - 6:52 am
Hi Dr. Muthen,

Thanks for your response. I am a little confused, though, about implementing the cutpoint code. My covariate or grouping variable, is defined by the code in Example 8.13 for a binary covariate. I have under the classes command: cg(2). How does this mesh with using the covariate, g, as a continuous variable with one cutpoint at 1 standard deviation above the mean?

Thanks,

Mary
 Bengt O. Muthen posted on Wednesday, April 18, 2018 - 3:52 pm
Then see the MonteCarlo version of ex 8.13 on our website at

http://www.statmodel.com/usersguide/chapter8.shtml
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: