Using Sigb PreviousNext
Mplus Discussion > Multilevel Data/Complex Sample >
Message/Author
 Anonymous posted on Sunday, October 16, 2005 - 4:37 pm
I have a quick question on using SIGB. I want to use the data that have been saved in SIGB to work with the between level portion of my multilevel SEM. The problem that I am having is accessing the sigb. My output is showing me that the information has been saved in the proper format and in the file that I had designated.

How would I set up the data and format statements to use the sigb information? I tried the usual methods as follows:

Data: File is c:\warner.dat;
Format is 26E15.8;

I have 26 variables that are at level-2. The manual on page 460 says that the data are saved in free format as E15.8. So, I used this format. I also tried this without the format statement, and I still got an error message. So, how would I use the sigb information?

Thank you
 Anonymous posted on Sunday, October 16, 2005 - 5:28 pm
I was able to figure out the problem. Thank you.
 Marco Haferburg posted on Tuesday, December 06, 2005 - 12:41 am
Hello Dr. Muthén,

I'am conducting several MFA, following the steps you described in your 1994-paper. To create the SIG-W- and SIG-B-matrices, I use the basic-option of type=twolevel with MLR. However, trying the same analysis without the basic-option, there's no convergence due to a non-pos between cov-matrix (some items have very little variance on the between-level).

My question is whether the estimated SIB-W and SIG-B created by the basic-run is trustworthy for the initial steps you described in your paper? (Anyway, in the last step, I had to eliminate the items with low between-variance from the between-part of the model in order to get a reasonable fit)

Many thanks!
 Linda K. Muthen posted on Tuesday, December 06, 2005 - 8:33 am
You don't say what your MODEL command is when you run the model without TYPE=BASIC; It is likely that the model you are specifying is not appropriate for the data.

If there are no warnings with TYPE=BASIC; the matrices created should be trustworthy.
 Marco Haferburg posted on Tuesday, December 06, 2005 - 9:34 am
I didn't specify a model with type=basic at all, but there are also no warnings.
 Linda K. Muthen posted on Tuesday, December 06, 2005 - 10:37 am
You should not specify a model with TYPE=BASIC. If there were no warnings, then the matrices should be fine.

When you ran the model without TYPE=BASIC, what model did you specify?
 Marco Haferburg posted on Wednesday, December 07, 2005 - 1:29 am
Hello Linda,

I tried a few options and experienced the following:
- With "type=twolevel basic" and no specified model, it was always possible to create the SIG-B and SIB-W
- Without the basic-option, the created files never contained any data, they were simply empty (no matter whether I specified a model or not)
- Convergence-problems didn´t occur with the basic-option; without the basic-option, it was sometimes necessary to define some items with low ICC as within in order to get a result, but the saved SIB-B/SIB-W were always empty

Since there were no warnings, I will work with the matrices created by the basic runs. Many thanks,
Marco
 Linda K. Muthen posted on Wednesday, December 07, 2005 - 6:03 am
Are you using Version 3.13? If not, you should download it. Otherwise, you need to send your various inputs, data, outputs, and license number to support@statmodel.com so I can see what is happening.
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: