Timeline for MCMC on a bounded parameter space?
Current License: CC BY-SA 3.0
9 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Oct 28, 2013 at 19:29 | comment | added | guy | @Cupitor I don't see him saying that. I think the implication is that Gabriel was doing Metropolis-within-Gibbs. | |
| Oct 28, 2013 at 17:52 | vote | accept | Cupitor | ||
| Oct 28, 2013 at 16:01 | comment | added | Cupitor | @guy, but according to the discussion on xian's page(above link by Zen), it looks like that Gibbs has a superiority without mentioning any reason though! | |
| Oct 27, 2013 at 23:30 | comment | added | guy | If MH proposes something outside of the parameter space, the acceptance probability is just set to $0$ and everything works fine. I think MH just interprets $0/0$ as $0$ (a manifestation of $0 \cdot \infty = 0$ in measure theory). | |
| Oct 27, 2013 at 23:05 | answer | added | jbowman | timeline score: 34 | |
| Oct 27, 2013 at 22:00 | comment | added | Cupitor | @Zen, I am not quite sure but the answer suggested by Xian is that to subsample, but instead of using MH, to use Gibbs sampler and reiterate if one of the values of a dimension exceeds the boundary, am I right? | |
| Oct 27, 2013 at 21:30 | comment | added | Zen | Take a look: xianblog.wordpress.com/tag/constrained-parameter-spaces | |
| Oct 27, 2013 at 21:09 | history | edited | Cupitor | CC BY-SA 3.0 |
edited title
|
| Oct 27, 2013 at 20:07 | history | asked | Cupitor | CC BY-SA 3.0 |