Bayesian statistics is a completely different paradigm for thinking about statistics. It involves making a complete probability model for what you think did happen, rather than constructing a framework to rule out things that didn’t happen (e.g., this result arising by chance). It has important advantages (conceptual simplicity, power, ability to combine information from different sources), and important disadvantages (requires more assumptions, answers may be difficult to find).
As a scientist you are probably well advised to take a pragmatic approach by using whatever statistical technique seems best suited to your problem. You are definitely well advised to make a good-faith effort to understand the foundations (particularly the philosophical foundations) of the methods you are using (as well as the methods your colleagues are using, since you will have to be able to evaluate these intelligently.
Lectures
Introductory
Discussion
“The first data analysis should be journalistic”: Edwards 1996 Ecological Applications
“Should ecologists become Bayesians?”: strong anti-Bayesian view, Dennis 1996 “Ecological Applications” (“Bayesianism means never having to say you’re wrong”; “[Bayesianism equals scientific relativism, which is … ] a sort of intellectual Calvinball”)
Translating Probability Density Functions: From R to BUGS and Back Again, LeBauer et al. 2013
Software
Example from an earlier class
Previous lectures
Install JAGS (and either rjags or R2jags).
Use jags to fit a Bayesian model to your data, in some way
that at least roughly makes sense. Discuss your prior assumptions, and
compare your simple fit to an analogous frequentist fit.
