Thanks to all for the helpful references on MCMC. These should keep me busy!
There's quite an interesting discussion of the coding of a Gibbs sampler at http://darrenjw.wordpress.com/2011/07/16/gibbs-sampler-in-various-languages-revisited/ along with timings of various implementations (R, C + GSL, python, etc.). I've implemented Darren Wilkinson's test case as a libgretl program (very fast) and also as a gretl script (much slower than C, but faster than python and a good deal faster than R). I'm currently experimenting to see how much of a speed-up might be available for the hansl version if there were a suitable dedicated command. -- Allin Cottrell Department of Economics Wake Forest University, NC
