| Message/Author |
|
|
| Anonymous posted on Thursday, June 23, 2005 - 12:03 pm
|
|
|
Acutally my data is collected from stratified sampling and there are a sampling weights variable(for unequal selection probability) and variables for Primary sampling unit and strata. I want to fit SEM model with my data.
My question is
"If I have the weight variables(BRR,Balanced Repeated Replication. Actually I have 88 BRR variables), how I can reflect this information in SEM? When I use Weight option, Cluster option and Stratification option in Variable command,
the standard error is based on Taylor series expansion. I know that the standard errors based on BRR usually are more reliable than the standard errors based on Taylor series expansion." |
|
|
|
|
| If you have 88 weight variables, this cannot be handled in Mplus. |
|
|
| Tom Munk posted on Monday, August 29, 2005 - 8:13 am
|
|
|
I have a situtation similar to that mentioned above, but I'm wondering if external Monte Carlo can help.
BRR (mentioned above) is a relative of Jackknife (my situation). In both cases, there are multiple weight variables because the analysis is intended to be run multiple times.
In my NAEP data, I have 62 replicate weights. I'm intended to run the data once with base weights to get parameter estimates, then 62 more times -- each time the same analysis with a different weight variable (sschwt01-62). The appropriate variance estimate is the sum of the squared differences of the sixty-two parameter estimates from the baseline parameter estimate. This can be used for significance tests.
I know I can use External Monte Carlo to run all sixty-three models and generate useful output if I first create sixty-three different datasets -- but the replicate weights I've described are designed to make this much simpler. I should be able to use the same dataset, simply changing the weight variable with each run. Does MPLUS provide any way to automate these 63 runs?
Thanks,
Tom |
|
|
|
|
| I can't think of any way to do this without having separate data sets. External Monte Carlo uses the same input and separate data sets but doesn't have different inputs. Alternatively, you could create different inputs and run them using a batch file in DOS. |
|
|
| Tom Munk posted on Monday, August 29, 2005 - 9:10 am
|
|
|
| That sounds interesting. Where might I find an example of a batch file creating multiple MPLUS runs? |
|
|
|
|
You need to create an ASCII file that contains the following:
call Mplus run1.inp
call Mplus run2.inp
call Mplus run3.inp
where run1.inp, run2.inp, and run3.inp are Mplus input files.
Name it run.bat or something ending in bat. Then go to the DOS prompt and type run if it is named run.bat and hit enter. |
|
|
| Tom Munk posted on Monday, August 29, 2005 - 9:48 am
|
|
|
| Is there any way short of hand-entry into an excel spreadsheet to capture and operate on the output files? |
|
|
|
|
| Mplus saves the results of each run. See the RESULTS option of the SAVEDATA command. |
|
|
|
|
| I'm planning to do secondary data analysis with data from nces that were collected with a complex sampling design. The dataset includes sampling weights and replicate weights to conduct JK2 analyses for computing adjusted standard errrors in programs like AM and WesVar. The dataset also includes a stratification variable and a cluster variable to account for the complex sampling design. Can you provide some guidance on conducting a CFA (and later an SEM) with both categorical and continuous indicators using the appropriate weights? |
|
|
|
|
| Mplus can handle sampling weights, clustering, and stratification. See the complex survey data features described in the user's guide. Mplus does not handle replicate weights. |
|
|
| Maja Cambry posted on Wednesday, May 02, 2007 - 9:28 am
|
|
|
I tried to conduct a simple regression using the sampling weight, cluster variable, and stratification variables that are available in my dataset. I get the following error message:
*** ERROR
Each stratum must contain unique cluster IDs.
Clusters are not nested within strata.
I then generated unique cluster ids for each stratum, but still had the same error message. Any clarification on this error message would help. Thanks,
M |
|
|
|
|
| Please send your input, data, output, and license number to support@statmodel.com. If you generated unique clusters, you would not get this error message. |
|
|
|
|
Dear Dr. Muthen,
I am trying to start multiple MPlus-runs using a batch file as suggested above (August 29, 2005). When I execute this file, the MPlus window opens, showing the first input file specified in the run.bat, but not processing anything. Once I close the MPlus window, another instance opens, showing the second input file, so per se the batch seems to work. What do I need to change in order to have the specified input files executed successively?
Thank you very much for your help! |
|
|
|
|
| Please send the relevant files and your license number to support@statmodel.com so I can see what the problem is. |
|
|
|
|
| I have to apologize, apparently I specified the wrong MPlus file in the batch file, everything works just fine now. Thank you nevertheless. |
|
|
| Back to top |
