pathikrit-bhattacharya/MCMC_fortran_library: FortMcMc_v1

Suite of adaptive MCMC routines in Fortran: The MCMC algorithm used here was based on the Metropolis-Hastings (MH) algorithm [Metropolis et al., 1953; Hastings, 1970] and makes use of the strategy of sampling the posterior using Bayes' theorem. The code is designed for fast convergence on correlated parameter spaces by adaptively learning the parameter covariance of the sampled region. The proposal jumps are then planned based on this covariance structure with step size prefixed according to some linear scaling of the covariance matrix. The scaling used is exact for large dimensional spaces as long as the moves of the proposals are diffusive. The document 'Description of using the MCMC code for a rate-state inverse problem.pdf' (which is really the Supplementary Material from Bhattacharya et al. [2015]) explains many of the working details of the code. The code also has options for setting a-priori constraints on any subset of the parameters. There are also options for running non-adaptive chains with no covariance structures and those with pre-defined covariance structures.