Hi all,
I have been working with student-t linear models, trying to estimate
degrees of freedom.
I used all non informative priors for linear parameters and error
variance (precision).
for degrees of freedom, I specified the following prior:
df<-1/df
id~dunif (0,0.5) # to be sure thar df are at least 2
When I run the model I detect convergence problems, specially in some
"estimable" functions of linear parameters, moreover, the posterior
distribution of df has mean 2 with variance close to zero.
I also tried a uniform prior on df:
df~dunif(a,b)
with different values of b: 30,100 and a: 2,3,5 and obtained similar
results: the sampling of df parameter get stucked in the lower bound
allowed by its prior.
The dataset I am analyzing has a sligthly thick tailed distribution, and
residual analysis reveals that maybe some value between 8 and 15 is
plausible for df.
I simmulated a simple dataset (df=8, mu=100, tau=0.1 and N=1000) and
I've winded up with convergence problems, but in the opposite direction:
df goes to a bigger and bigger value and has a huge posterior variance.
I would appreciate very much if someone has comments or observations.
thank you all
--
=============================
Juan Pedro Steibel
Graduate Student
Department of Animal Science
Michigan State University
1261 Anthony Hall
East Lansing, MI
48823 USA
Phone: 1-517-432-0671
E-mail: [log in to unmask]
=============================
-------------------------------------------------------------------
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
|