mcmc

Monte-Carlo Markov Chain methods.

run_mcmc(spectro, positions, simcfg, ...[, ...])	Monte-Carlo Markov Chain exploration of parameter space, using emcee.
plot_mcmc_chains(chains, parameters)	Plot MCMC-chains (nwalkers, nsteps, ndim).
plot_mcmc_corner(chains, parameters[, burnin])	corner plot of MCMC parameters.
mcmc_best_params(chains, parameters[, burnin])	Compute best parameter estimates and assymetric 1-sigma errors from marginalized distributions.

spectrogrism.mcmc.run_mcmc(spectro, positions, simcfg, optparams, lnprior, modes=None, nwalkers=10, nsteps=500, outfile='chains.dat')

Monte-Carlo Markov Chain exploration of parameter space, using emcee.

Warning

very preliminary implementation

Parameters:

• spectro (Spectrograph) – spectrograph
• positions (DetectorPositions) – target positions
• simcfg (SimConfig) – simulation configuration
• optparams (list) – optical parameters to be probed
• lnprior (function) – log-likelihood prior function
• modes (list) – adjusted observing modes (default: simulated modes)
• nwalkers (int) – the number of walkers will be 2*len(optparams)*nwalkers
• nsteps (int) – number of MCMC-steps
• outfile (str) – incremental file output

Returns:

Monte-Carlo Markov Chains array (nwalkers, nsteps, ndim)

Return type:

emcee.EnsembleSampler.chain

Raises:

KeyError – unknown optical parameter

Reference: Foreman-Mackey et al. 2012

spectrogrism.mcmc.plot_mcmc_chains(chains, parameters)

Plot MCMC-chains (nwalkers, nsteps, ndim).

spectrogrism.mcmc.plot_mcmc_corner(chains, parameters, burnin=0)

corner plot of MCMC parameters.

Reference: Foreman-Mackey 2016

spectrogrism.mcmc.mcmc_best_params(chains, parameters, burnin=0)

Compute best parameter estimates and assymetric 1-sigma errors from marginalized distributions. Return {name: (median, low_err, high_err)}.