[astro-ph/0405462] Fast and reliable MCMC for cosmological p
Posted: September 24 2004
The result for scaling the proposal density with dimension given in the paper appears to be generally useful and correct. Two comments:
1. There are some potential problems using the autocorrelation function to assess chain efficiency: it is quite possible for the autocorrelation to be small but the separation between effectively independence samples to be large. Jo has confirmed that similar problems can happen with the spectral estimator.
2. Results are restricted to using N-D Gaussians as the proposal density. In high-N (N>2) this is probably rather un-robust, and it may be better to use a lower dimensional proposal.
I've written up these comments in detail in the cosmomc notes
http://cosmologist.info/notes/cosmomc.ps.gz
Similar comments apply to [arxiv]astro-ph/0310723[/arxiv]
1. There are some potential problems using the autocorrelation function to assess chain efficiency: it is quite possible for the autocorrelation to be small but the separation between effectively independence samples to be large. Jo has confirmed that similar problems can happen with the spectral estimator.
2. Results are restricted to using N-D Gaussians as the proposal density. In high-N (N>2) this is probably rather un-robust, and it may be better to use a lower dimensional proposal.
I've written up these comments in detail in the cosmomc notes
http://cosmologist.info/notes/cosmomc.ps.gz
Similar comments apply to [arxiv]astro-ph/0310723[/arxiv]