"""Plot pointwise elpd estimations of inference data."""
import numpy as np
from ..rcparams import rcParams
from ..stats import _calculate_ics
from ..utils import get_coords
from .plot_utils import format_coords_as_labels, get_plotting_function
[docs]
def plot_elpd(
compare_dict,
color="C0",
xlabels=False,
figsize=None,
textsize=None,
coords=None,
legend=False,
threshold=None,
ax=None,
ic=None,
scale=None,
var_name=None,
plot_kwargs=None,
backend=None,
backend_kwargs=None,
show=None,
):
r"""Plot pointwise elpd differences between two or more models.
Pointwise model comparison based on their expected log pointwise predictive density (ELPD).
Notes
-----
The ELPD is estimated either by Pareto smoothed importance sampling leave-one-out
cross-validation (LOO) or using the widely applicable information criterion (WAIC).
We recommend LOO in line with the work presented by [1]_.
Parameters
----------
compare_dict : mapping of {str : ELPDData or InferenceData}
A dictionary mapping the model name to the object containing inference data or the result
of :func:`arviz.loo` or :func:`arviz.waic` functions.
Refer to :func:`arviz.convert_to_inference_data` for details on possible dict items.
color : str or array_like, default "C0"
Colors of the scatter plot. If color is a str all dots will have the same color.
If it is the size of the observations, each dot will have the specified color.
Otherwise, it will be interpreted as a list of the dims to be used for the color code.
xlabels : bool, default False
Use coords as xticklabels.
figsize : (float, float), optional
If `None`, size is (8 + numvars, 8 + numvars).
textsize : float, optional
Text size for labels. If `None` it will be autoscaled based on `figsize`.
coords : mapping, optional
Coordinates of points to plot. **All** values are used for computation, but only a
subset can be plotted for convenience. See :ref:`this section <common_coords>`
for usage examples.
legend : bool, default False
Include a legend to the plot. Only taken into account when color argument is a dim name.
threshold : float, optional
If some elpd difference is larger than ``threshold * elpd.std()``, show its label. If
`None`, no observations will be highlighted.
ic : str, optional
Information Criterion ("loo" for PSIS-LOO, "waic" for WAIC) used to compare models.
Defaults to ``rcParams["stats.information_criterion"]``.
Only taken into account when input is :class:`arviz.InferenceData`.
scale : str, optional
Scale argument passed to :func:`arviz.loo` or :func:`arviz.waic`, see their docs for
details. Only taken into account when values in ``compare_dict`` are
:class:`arviz.InferenceData`.
var_name : str, optional
Argument passed to to :func:`arviz.loo` or :func:`arviz.waic`, see their docs for
details. Only taken into account when values in ``compare_dict`` are
:class:`arviz.InferenceData`.
plot_kwargs : dicts, optional
Additional keywords passed to :meth:`matplotlib.axes.Axes.scatter`.
ax : axes, optional
:class:`matplotlib.axes.Axes` or :class:`bokeh.plotting.Figure`.
backend : {"matplotlib", "bokeh"}, default "matplotlib"
Select plotting backend.
backend_kwargs : dict, optional
These are kwargs specific to the backend being used, passed to
:func:`matplotlib.pyplot.subplots` or :class:`bokeh.plotting.figure`.
For additional documentation check the plotting method of the backend.
show : bool, optional
Call backend show function.
Returns
-------
axes : matplotlib_axes or bokeh_figure
See Also
--------
plot_compare : Summary plot for model comparison.
loo : Compute Pareto-smoothed importance sampling leave-one-out cross-validation (PSIS-LOO-CV).
waic : Compute the widely applicable information criterion.
References
----------
.. [1] Vehtari et al. (2016). Practical Bayesian model evaluation using leave-one-out
cross-validation and WAIC https://arxiv.org/abs/1507.04544
Examples
--------
Compare pointwise PSIS-LOO for centered and non centered models of the 8-schools problem
using matplotlib.
.. plot::
:context: close-figs
>>> import arviz as az
>>> idata1 = az.load_arviz_data("centered_eight")
>>> idata2 = az.load_arviz_data("non_centered_eight")
>>> az.plot_elpd(
>>> {"centered model": idata1, "non centered model": idata2},
>>> xlabels=True
>>> )
.. bokeh-plot::
:source-position: above
import arviz as az
idata1 = az.load_arviz_data("centered_eight")
idata2 = az.load_arviz_data("non_centered_eight")
az.plot_elpd(
{"centered model": idata1, "non centered model": idata2},
backend="bokeh"
)
"""
try:
(compare_dict, _, ic) = _calculate_ics(compare_dict, scale=scale, ic=ic, var_name=var_name)
except Exception as e:
raise e.__class__("Encountered error in ic computation of plot_elpd.") from e
if backend is None:
backend = rcParams["plot.backend"]
backend = backend.lower()
numvars = len(compare_dict)
models = list(compare_dict.keys())
if coords is None:
coords = {}
pointwise_data = [get_coords(compare_dict[model][f"{ic}_i"], coords) for model in models]
xdata = np.arange(pointwise_data[0].size)
coord_labels = format_coords_as_labels(pointwise_data[0]) if xlabels else None
if numvars < 2:
raise ValueError("Number of models to compare must be 2 or greater.")
elpd_plot_kwargs = dict(
ax=ax,
models=models,
pointwise_data=pointwise_data,
numvars=numvars,
figsize=figsize,
textsize=textsize,
plot_kwargs=plot_kwargs,
xlabels=xlabels,
coord_labels=coord_labels,
xdata=xdata,
threshold=threshold,
legend=legend,
color=color,
backend_kwargs=backend_kwargs,
show=show,
)
plot = get_plotting_function("plot_elpd", "elpdplot", backend)
ax = plot(**elpd_plot_kwargs)
return ax
