Module Documentation

emcwrap.sampler

emcwrap.sampler.get_prior_sample(frozen_prior, nsamples, check_func=False, seed=None, mapper=<class 'map'>, filterwarnings='error', max_attempts=10, debug=False, verbose=True)

Get a sample of size nsamples from the prior distribution.

emcwrap.sampler.mcmc_summary(chain, lprobs, priors, acceptance_fraction=None, out=<built-in function print>, **args)
emcwrap.sampler.run_mcmc(lprob, nsteps, p0=None, moves=None, priors=None, prior_transform=None, backend=None, update_freq=False, resume=False, pool=None, report=None, description=None, temp=1, maintenance_interval=False, verbose=True, **kwargs)

Run the emcee sampler.

emcwrap.moves

emcwrap.moves.ADEMove

alias of DIMEMove

class emcwrap.moves.DIMEMove(*args: Any, **kwargs: Any)

Bases: RedBlueMove

A proposal using adaptive differential-independence mixture ensemble MCMC.

This is the Differential-Independence Mixture Ensemble proposal as developed in Ensemble MCMC Sampling for DSGE Models (previousy ADEMC).

Parameters

  • sigma (float, optional) – standard deviation of the Gaussian used to stretch the proposal vector.

  • gamma (float, optional) – mean stretch factor for the proposal vector. By default, it is \(2.38 / \sqrt{2\,\mathrm{ndim}}\) as recommended by ter Braak (2006).

  • aimh_prob (float, optional) – probability to draw an adaptive independence Metropolis Hastings (AIMH) proposal. By default this is set to \(0.1\).

  • df_proposal_dist (float) – degrees of freedom of the multivariate t distribution used for AIMH proposals. Defaults to \(10\).

get_proposal(x, dummy, random)
propose(model, state)
setup(coords)
emcwrap.moves.mvt_sample(df, mean, cov, size, random)

Sample from multivariate t distribution

For reasons beyond my understanding, the results from random.multivariate_normal with non-identity covariance matrix are not reproducibel across architecture. Since scipy.stats.multivariate_t is based on numpy’s multivariate_normal, the workaround is to crochet this manually. Advantage is that the scipy dependency drops out.

emcwrap.tools

emcwrap.tools.find_mode_simple(lprob, init, frozen_prior, sd=True, verbose=False, **kwargs)
emcwrap.tools.get_bijective_prior_transformation(funcs_con, funcs_re)

Get the bijective prior transformation function.

emcwrap.tools.get_log_prior(frozen_prior)

Get the log-prior function.

emcwrap.tools.get_prior(prior, verbose=False)

Compile prior-related computational objects from a list of priors.

emcwrap.tools.load_backend(backend)

just a shortcut

emcwrap.tools.remove_backend(backend)

just a shortcut

emcwrap.tools.rm_backend(backend)

just a shortcut

emcwrap.tools.save_to_backend(backend, content)

emcwrap.plots

emcwrap.plots.fast_kde(x, bw=4.5)

A fft-based Gaussian kernel density estimate (KDE) The code was adapted from https://github.com/mfouesneau/faststats

Parameters

  • x (Numpy array or list) –

  • bw (float) – Bandwidth scaling factor for the KDE. Should be larger than 0. The higher this number the smoother the KDE will be. Defaults to 4.5 which is essentially the same as the Scott’s rule of thumb (the default rule used by SciPy).

Returns

  • density (A gridded 1D KDE of the input points (x))

  • xmin (minimum value of x)

  • xmax (maximum value of x)

emcwrap.plots.get_axis(ax, default_rows, default_columns, **default_kwargs)

Verifies the provided axis is of the correct shape, and creates one if needed.

Parameters

  • ax – matplotlib axis or None

  • default_rows – int, expected rows in axis

  • default_columns – int, expected columns in axis

  • **default_kwargs – keyword arguments to pass to plt.subplot

Returns

axis, or raises an error

emcwrap.plots.kdeplot_op(ax, data, bw, prior=None, prior_alpha=1, prior_style='--')

Get a list of density and likelihood plots, if a prior is provided.

emcwrap.plots.plot_posterior_op(trace_values, ax, bw, prior, kde_plot=False, point_estimate='mean', round_to=3, alpha_level=0.05, ref_val=None, rope=None, text_size=16, display_additinal_info=False, **kwargs)

Artist to draw posterior.

emcwrap.plots.posteriorplot(trace, varnames=None, tune=0, figsize=None, plots_per_fig=4, text_size=None, ropep=None, ref_val=None, round_to=3, alpha_level=0.05, bw=4.5, ax=None, **kwargs)
emcwrap.plots.scale_text(figsize, text_size)

Scale text to figsize.

emcwrap.plots.traceplot(trace, varnames, tune, figsize=None, plots_per_fig=3, frozen_prior=None, draw_each_trace=True, bw=4.5, text_size=None, **kwargs)

emcwrap.stats

emcwrap.stats.calc_min_interval(x, alpha)

Internal method to determine the minimum interval of a given width

Assumes that x is sorted numpy array.

emcwrap.stats.gelman_rubin(sample)

The Gelman-Rubin (1992) statistics.

emcwrap.stats.integrated_time(x)
emcwrap.stats.mc_error(x)
emcwrap.stats.mdd_harmonic_mean(chain, lprobs, pool=None, alpha=0.05, verbose=False, debug=False)

Approximate the marginal data density useing modified harmonic mean.

emcwrap.stats.mdd_laplace(chain, lprobs, calc_hess=False)

Approximate the marginal data density useing the LaPlace method.

emcwrap.stats.summary(priors, store, pmode=None, bounds=None, alpha=0.1, top=None, show_prior=True)

emcwrap.dists

class emcwrap.dists.InvGammaDynare(*args: Any, **kwargs: Any)

Bases: rv_continuous

name = 'inv_gamma_dynare'
emcwrap.dists.inv_gamma_spec(mu, sigma)