 |
 |
Charlotte, Sometimes I have had convergence problems because of the way that I parameterized. For example, is it conceivable that the precision is very large (standard deviation, or perhaps coefficient of variation, near zero)? In that case, the posterior distribution for the precision could be spread out all over the extreme part of the axis, but the posterior distribution for the standard deviation could converge nicely to some distribution near zero. The effect is exaggerated with a nearly improper prior such as gamma(0.001, 0.001). Cory Atwood ----- Original Message ----- From: "Charlotte Gary" <[log in to unmask]> To: <[log in to unmask]> Sent: Monday, May 30, 2011 6:13 AM Subject: [BUGS] uniformative prior for precision: gamma(0.5,0.5) vs. gamma(0.001,0.001) > Hi all, > > here is my issue. I run two different hierachical bayesian models, > the first using gamma(0.5,0.5) as prior for precision and the second > using gamma(0.001,0.001). > I let these models run for a quite high number of iterations (one > million iterations for both). > At the end of this process. The model with gamma(0.5,0.5) nicely > converged (the History plots of the average posterior parameter > estimates are quite "stable" after 150k iterations). However, the > model with gamma(0.001,0.001) shows some convergence problems for some > parameters. > > Any idea of why this is happening? > > Honestly I am quite happy of the results obtained with gamma(0.5,0.5) > they make completly sense. However, it seems that gamma(0.001,0.001) > is a quite common prior. I was not able to find good references for > gamma(0.5,0.5). Have u seen previous work using this values? Would you > trust these values? > > any comment would be strongly appreciated! > > Thank you in advance for you help! > > ------------------------------------------------------------------- > 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 > > ------------------------------------------------------------------- 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