arviz_plots.plot_energy

Contents

arviz_plots.plot_energy#

arviz_plots.plot_energy(dt, kind=None, threshold=0.3, sample_dims=None, plot_collection=None, backend=None, labeller=None, aes_by_visuals=None, visuals=None, stats=None, **pc_kwargs)[source]#

Plot energy distributions and bfmi from gradient-based algorithms.

Generate a figure with two plots: On the left the Bayesian Fraction of Missing Information (BFMI) per chain, values below the threshold indicate poor exploration of the energy distribution. On the right, the marginal energy distribution and the energy transition distribution. Ideally, these two distributions should overlap closely.

For details on BFMI and energy diagnostics see [1] for a more practical overview check the EABM chapter on MCMC diagnostic of gradient-based algorithms.

Parameters:
dtxarray.DataTree

sample_stats group with an energy variable is mandatory.

kind{“kde”, “hist”, “dot”, “ecdf”}, optional

How to represent the marginal density. Defaults to rcParams["plot.density_kind"]

thresholdfloat, default 0.3

Reference threshold for BFMI values, values below this indicate poor exploration of the energy distribution.

sample_dimssequence of str, optional

Dimensions to consider as sample dimensions when computing BFMI. Defaults to rcParams["data.sample_dims"]

plot_collectionPlotCollection, optional
backend{“matplotlib”, “bokeh”, “plotly”}, optional
labellerlabeller, optional
aes_by_visualsmapping of {strsequence of str}, optional

Mapping of visuals to aesthetics that should use their mapping in plot_collection when plotted. Valid keys are the same as for visuals.

visualsmapping of {strmapping or bool}, optional

Valid keys are:

  • dist -> depending on the value of kind passed to:

    • “kde” -> passed to line_xy

    • “ecdf” -> passed to ecdf_line

    • “hist” -> passed to step_hist

    • “dot” -> passed to scatter_xy

  • title -> passed to labelled_title

  • legend -> passed to arviz_plots.PlotCollection.add_legend

  • remove_axis -> not passed anywhere, can only be False to skip calling this function

  • title -> passed to labelled_title

  • bfmi_points -> passed to scatter_xy for BFMI scatter plot

  • ylabel -> passed to labelled_y for BFMI column y-axis label

  • face -> visual that fills the area under the energy distributions.

    Defaults to True. Depending on the value of kind it is passed to:

    • “kde” or “ecdf” -> passed to fill_between_y

    • “hist” -> passed to hist

    • dot -> ignored

statsmapping, optional

Valid keys are:

  • dist -> passed to kde, ecdf, …

**pc_kwargs

Passed to arviz_plots.PlotCollection.wrap

Returns:
PlotCollection

References

[1]

Betancourt. Diagnosing Suboptimal Cotangent Disintegrations in Hamiltonian Monte Carlo. (2016) https://arxiv.org/abs/1604.00695

Examples

Plot an energy plot using ecdf for the energy distributions.

>>> from arviz_plots import plot_energy, style
>>> style.use("arviz-variat")
>>> from arviz_base import load_arviz_data
>>> data = load_arviz_data('non_centered_eight')
>>> plot_energy(data, kind="ecdf")
../../_images/arviz_plots-plot_energy-1.png

Plot transition and marginal energy distributions

Energy

Energy
Contents