Hi
One month ago I posted two questions to the list. Thanks for the answers. I realised that perhaps it was not the most understandable questions in the world, but below is what was important for me.
>Can you somehow restrict your parameters in the dispersion matrix in MVN to be equal
>e.g. equal variances or equal covariances in the general n x n case. In the 2 x 2 case you could use the trick
>with the conditional formulation, but is there an easier solution when going to higher dimensions?
There is a new example on the winbugs site, where an autoregression is modelled using the fact that the inverse covariance matrix also has a simple form:
http://www.mrc-bsu.cam.ac.uk/bugs/winbugs/examples/pigweights.txt
Also the covariance matrix with a equal variance and equal off-diagonal entries have a inverse covariance matrix of a rather simple form see. eg. Seber: multivariate observations pp 520, else I think it is possible to use different random effects as residuals on different levels imposing a positive correlation see eg. Congdon: Bayesian statistical modelling pp. 381 (example 8.4)
>Also, how sentisive is the dispersion parameters from the starting values in the 'non-informative' Wishart prior with n degrees >>of freedom, looking at dispersion matrices of relative high dimensions e.g. n=10. Should one make an efffort to have good >>>starting values from e.g. SAS proc mixed/glimmix.
Yes, with sparse data good starting values can be beneficial...
best regards
Soren
-------------------------------------------------------------------
This list is for discussion of modelling issues and the BUGS software.
For help with crashes and error messages, first mail [log in to unmask]
To mail the BUGS list, mail to [log in to unmask]
Before mailing, please check the archive at www.jiscmail.ac.uk/lists/bugs.html
Please do not mail attachments to the list.
To leave the BUGS list, send LEAVE BUGS to [log in to unmask]
If this fails, mail [log in to unmask], NOT the whole list
|