www.cmm.bristol.ac.uk/forum

It's possible that your data isn't sorted in the order of the model hierarchy. MLwiN generally uses the fact that the unit number has changed to determine that it is the next unit so if you had the same unit number twice, but with a different one in between it would be counted as two different units.

This is a usually caused by a Stata limit, see http://www.cmm.bristol.ac.uk/forum/viewtopic.php?f=3&t=176 for a possible workaround. Note, you should not longer need to increase the MLwiN maximum number of explanatory variables as indicated in the solution because runmlwin now sets this based on ...

There are two likely causes of this error message:

1) You don't have enough worksheet space allocated (It needs at least the length of the source column to be available), in which case you can fix it by increasing the worksheet size.

or

2) The source column contains no data.

Yes, you should receive a link to the download from info-cmm@bristol.ac.uk.

In order to use Realcom-Impute you need to first install the Matlab runtime that matches the version that it was compiled with (currently 2009a). You can obtain this by following the instructions given in http://www.bristol.ac.uk/cmm/software/realcom/#matlab.

It sounds like you may have set a worksheet size that is larger than MLwiN is able to allocate. I would suggest that you initially try the steps outlined in http://www.bris.ac.uk/cmm/software/supp ... ml#default to see whether that fixes the problem.

As long as the chains have converged to a stable distribution I believe that the DIC should be comparable. You can find more information on how the DIC is calculated in section 2.2 of the MCMC manual.

I am not sure what is causing this specific error, however if you can provide an example worksheet that exhibits that behaviour I will look into it.

Please note that we don't recommend using (R)IGLS for logistic regression for the reasons given in http://www.bristol.ac.uk/cmm/software/support ...

900,000Kcells works out as about 3.4Gb which it is very unlikely that MLwiN can allocate in one contiguous block. It is possible that the calculation runmlwin is doing to work out the space needed is wrong, in which case you can try using the mlwinsettings(size()) option with lower values. The ...

Multiple membership with sampling weights is not possible with MCMC in MLwiN. You can set up multiple membership models with (R)IGLS using special commands (see chapter 19 in the MLwiN manual: http://www.bristol.ac.uk/cmm/software/mlwin/download/mlwin-userman-09.pdf ). If you set it up this way then ...