Dear all,
I'm fitting a model using MCMC estimation. After running a 20k chain, I stored the estimates and saved the worksheet as usual; nothing wrong so far. Afterwards, I closed MLwiN and reopened the worksheet a few hours later and got the following error message:
"MCMC Error 0327: This variance structure cannot currently be fitted using MCMC (RETR)"
The message pops up before loading the worksheet and the only choice I'm given is to click OK. After that, MLwiN just freezes and the worksheet never loads.
Any ideas about why this happened?
Thanks in advance!
Best,
Patricio
MCMC error 0327
-
- Posts: 1384
- Joined: Mon Oct 19, 2009 10:34 am
Re: MCMC error 0327
I'm not sure why this error occurs in your case, it is possible that something wasn't saved properly, but I would need to see the worksheet to determine what. If you are able to send me a copy of this then I will attempt to investigate what is happening.
One thing to note is that if there are any issues loading the worksheet then MLwiN does not currently automatically open the Names Window, meaning that there is no visual cue to indicate that it has finished. It is therefore possible that loading the worksheet got further than it appeared. To check this you could try opening the Names or Data windows yourself after attempting to load the worksheet and see whether any data has been loaded.
One thing to note is that if there are any issues loading the worksheet then MLwiN does not currently automatically open the Names Window, meaning that there is no visual cue to indicate that it has finished. It is therefore possible that loading the worksheet got further than it appeared. To check this you could try opening the Names or Data windows yourself after attempting to load the worksheet and see whether any data has been loaded.
-
- Posts: 9
- Joined: Wed Nov 21, 2012 12:24 pm
Re: MCMC error 0327
Thanks, Chris
Due to University regulations, I'm not allowed to send you the worksheet. I'll attempt to fit the model in a small sample of the dataset to replicate the error; I'll try to send that to you as soon as possible. Just one thing: can I send it to your email instead of attaching it to this forum?
I've tried to fit this model many times and it always produces the same error. In the meantime, can I ask you: Are there any known issues with multivariate cross-classified models that might be related to this problem?
Another thing, when trying to open the worksheet and after clicking OK in the error message window, I cannot click anywhere, so I cannot try to load the names or equations window from the menu bar either. I have actually waited for at least an hour for the worksheet to load, perhaps I should keep it there for longer. Maybe I'll get lucky! I'll try it and let you know.
Best,
Patricio
Due to University regulations, I'm not allowed to send you the worksheet. I'll attempt to fit the model in a small sample of the dataset to replicate the error; I'll try to send that to you as soon as possible. Just one thing: can I send it to your email instead of attaching it to this forum?
I've tried to fit this model many times and it always produces the same error. In the meantime, can I ask you: Are there any known issues with multivariate cross-classified models that might be related to this problem?
Another thing, when trying to open the worksheet and after clicking OK in the error message window, I cannot click anywhere, so I cannot try to load the names or equations window from the menu bar either. I have actually waited for at least an hour for the worksheet to load, perhaps I should keep it there for longer. Maybe I'll get lucky! I'll try it and let you know.
Best,
Patricio
-
- Posts: 1384
- Joined: Mon Oct 19, 2009 10:34 am
Re: MCMC error 0327
Yes, you can send this file directly to me, you can find my email address on http://www.bris.ac.uk/cmm/team/.
I can't recall any issues with multivariate cross-classified models that might cause this issue, although it sounds like it may be a saving/loading issue and therefore might be less likely to have been encountered.
I can't recall any issues with multivariate cross-classified models that might cause this issue, although it sounds like it may be a saving/loading issue and therefore might be less likely to have been encountered.