www.cmm.bristol.ac.uk/forum

The number of cases reported here are for the initially loaded data, rather than each of the imputed datasets (which would all have complete data). This is because these are only loaded temporarily while the imputation analysis is being run, and the original data is restored once this is complete.

The "Cases in use" numbers reported in the equations window reflect the missing data in the currently loaded data. While the imputation analysis is run this is temporarily replaced with each of the imputed dataset from Realcom, however once the analysis is complete and the model results ar...

You can find the MCMC chains in the chains slot of the mlwinfitMCMC object returned from the runMLwiN function. An example of this would be as follows: # for runMLwiN function library(R2MLwiN) # for gelman.diag function library(coda) # load the data data(tutorial, package = "R2MLwiN") # ru...

Unfortunately I don't think that there is currently a straightforward way to do this. If you look at lines 410 and 535 of ( mcmcsum.ado ) you will see that the results displayed for the summaries and detail option are stored returned in an r-class after the command is run, although as you say this i...

It looks as if the estimates replay command does not carry across the display options set when running the command, however you can re-specify them when displaying the results, for example: . use http://www.bristol.ac.uk/cmm/media/runmlwin/bang, clear . logit use age, or Iteration 0: log likelihood ...

The files that you provided ran successfully in both MLwiN versions 3.04 and 3.05 for me. You do not have to open a dataset prior to running the macro as all the values used are simulated. I set the current directory to the location of the macro files via Options>Directories, and opened the simu.txt...

I am not sure where the dashes in that syntax are coming from. Would it be possible to post the generated script, and if possible the inputs you gave to MLPowSim to create it?

If you are running a binary logit model then the variance is fixed at (π^2)/3 (≈3.29) rather than being an estimated parameter, and this is why you cannot add this to your model.

See the following FAQ for more information about variances: Partitioning variation across levels

I showed this to George, but he wasn't able to say for sure. His advice was that unless the models ran too slowly or couldn't be fitted with mixed you should use that. If you can't do that then he suggests initially running a subsample or simplified version of your model using both mixed and your ma...

This forum is primarily for questions related to the "R2MLwiN" R package. You might get an answer if you ask about related packages, but there are often more appropriate places for these questions.