# Notes

• Added multiple chains, randomized starting points for each chain, and confirming Gelman-Rubin convergence criteron via autojags.
• Some edits to adapt the extraction of posteriors from the multiple chains.
• Stange behavior in jags.parallel SO/16723036, looks like an annoying bug.
• Ah-ha, and now a nice solution to this problem via do.call.

When running on farm, try to remember: (yes it’s in the run.sh script already).

module load gcc R jags

This appears to result in greater uncertainty:

Suggesting the earlier appearance of convergence occurs only thanks to starting conditions close to the optimum?

## Log

• parallel jags working now 05:53 pm 2013/05/23
• cleanup 04:50 pm 2013/05/23
• generic version of allen 04:09 pm 2013/05/23
• Revising MCMC iterations

• Added multiple chains, randomized starting points for each chain, and confirming Gelman-Rubin convergence criteron via autojags.
• Some edits to adapt the extraction of posteriors from the multiple chains.
• Stange behavior in jags.parallel SO/16723036, looks like an annoying bug. 04:05 pm 2013/05/23
• No reply from query to package maintainer, but the SO link now has a clever work-around to said annoying bug.

## Misc

• Grr, why is dcast not more intuitive to me. Thank you SO folks. SO/16722117