Warning signals package
Reviewing Numerical challenges for faster analysis on warningsignals
created a quick mcmcTools package so different packages can use my custom routines. Will have to explore some of the standard packages eventually, though most seem somewhat specialized (the Bayesian task-view is epic). bayesm looks promising, and LearnBayes package that accompanies Jim Albert’s book (Albert, unknown) might be worth a closer look.
Wrote mcmc script for warningsignals, much more debugging to handle out-of-bounds conditions then expected. Also switched to call parameters by position instead of name in linear_bifur_models. Debugging is slow with slow codes!
illustrating how increasing the sampling rate increases power to determine necessary monitoring scheme:
[flickr-gallery mode=“search” tags=“stochpop” min_upload_date=“2011-05-24 10:40:00” max_upload_date=“2011-05-24 16:15:00”]
- Need to start writing more informative commit messages and committing more logically connected blocks with git. Will perhaps go hand-in-hand with getting better with fugitive plugin for git commands in vim, for which vimcasts has a nice tutorial.
Meetings/Talks
Matt @ Wainwright lab meeting 2-4p
Morgan’s exit seminar. 4-5p
References
- Albert J (unknown). “Bayesian Computation With R.” Unknown.