Sampling

There are facilities for sampling models from an SLP. Furthermore the output can be displayed in postscript format, and a latex report can be generated. Sampling is not a part of the MCMC algorithms but it can be very useful in providing feedback for newly developed SLPs: in particular, about their digestibility by the software and the distributions they define. For the time being, and until we expand on this section, please look at the source files: auxil/n_samples.pl and auxil/n_samples_tex

A round-about way to sampling is to run MCMC with a constant ratio of likelihood that is equal to something substantially larger than 1. In this way, the chain will almost certainly consist of the proposed models. (i.e. the proposed model is always accepted.) Such a likelihood can be found in auxil/lhood_constant.pl. An example of this technique is in auxil/from_prior_ex/srun.pl.