New kids on the Bayesian block
I attended Daniel Turek’s talk at ISEC about NIMBLE a neat alternative to WinBUGS and JAGS. It is developed by Perry de Valpine’s group at Berkeley and ‘lets you use BUGS models natively in R, program functions that use them, and compile everything via C++ for faster computing’. I played around with NIMBLE a bit, here is an example of fitting a classic capture-recapture model to simulated data - thanks to Perry and Daniel for their help! NIMBLE seems a lot faster than its competitors, and much more flexible. I’ll continue my investigations with more complex models - stay tuned.