MCMC Error in CAR modelling
Posted: Thu Apr 15, 2010 10:32 am
Hi MLwiN forum users,
I am hoping that someone may be able to help me. I have been using MLwiN for the last couple of years to fit multi-level models. I am now interested in fitting spatial models.
I have been following the guidance on Chapter 16 of the MCMC estimation in MLwiN manual and managed to fit a spatial multiple membership model. However, I am keen to see if a CAR model is more appropriate for my problem but when I try to fit this model I get the error message:
MCMC Error 0243: The ith neighbouring unit for area j is not a known area(MCMC)
where i and j are not always the same when I try and fit the model. However, i is commonly 1. I have checked area j and neighbour i to see if this error makes sense and the neighbour ID appears appropriate, as does the area number. I tried looking for error 0243 using search engines but couldn't find anything referring to this. Has anyone come across this error? Am I just doing something stupid with my data? I am fairly new to spatial regression analysis so would appreciate any guidance. It may just be that my data set is too large to cope with this problem in MLwiN- I have about 6500 areas with up to a possible 39 neighbours.
Cheers,
Karen
I am hoping that someone may be able to help me. I have been using MLwiN for the last couple of years to fit multi-level models. I am now interested in fitting spatial models.
I have been following the guidance on Chapter 16 of the MCMC estimation in MLwiN manual and managed to fit a spatial multiple membership model. However, I am keen to see if a CAR model is more appropriate for my problem but when I try to fit this model I get the error message:
MCMC Error 0243: The ith neighbouring unit for area j is not a known area(MCMC)
where i and j are not always the same when I try and fit the model. However, i is commonly 1. I have checked area j and neighbour i to see if this error makes sense and the neighbour ID appears appropriate, as does the area number. I tried looking for error 0243 using search engines but couldn't find anything referring to this. Has anyone come across this error? Am I just doing something stupid with my data? I am fairly new to spatial regression analysis so would appreciate any guidance. It may just be that my data set is too large to cope with this problem in MLwiN- I have about 6500 areas with up to a possible 39 neighbours.
Cheers,
Karen