A reversible jump MCMC algorithm for hierarchical distance sampling models

As part of my PhD I developed Bayesian methods for analysing distance sampling data. Here, I combined the likelihood components from the detection and count models and used this integrated likelihood in a reversible jump (RJ) MCMC algorithm. Some of the benefits of this method are that it allows estimating all model parameters and their uncertainty simultaneously. Besides parameter space, model space is explored, hence model selection becomes part of the inference.

The motivating data set for this analysis was a point transect survey in the US where the interest was whether planting herbaceous buffers around agricultural fields had the desired effect of improving the habitat for birds. More on this can be found in my JABES paper or in Kristine Evans's Journal of Wildlife Management paper.