change kde limits and hdi computation

CHANGELOG.md

@@ -11,10 +11,14 @@
 ### Documentation
 
 ## v0.8.3 (2020 May 28)
+### New features
+* loo-pit plot. The kde is computed over the data interval (this could be shorter than [0, 1]). The hdi is computed analitically (#1215) 
+
 ### Maintenance and fixes
 * Restructured internals of `from_pymc3` to handle old pymc3 releases and
   sliced traces and to provide useful warnings (#1211)
 
+
 ## v0.8.2 (2020 May 25)
 ### Maintenance and fixes
 * Fixed bug in `from_pymc3` for sliced `pymc3.MultiTrace` input (#1209)

arviz/plots/backends/bokeh/loopitplot.py

@@ -1,10 +1,10 @@
 """Bokeh loopitplot."""
 import bokeh.plotting as bkp
+from bokeh.models import BoxAnnotation
 import numpy as np
 
 from . import backend_kwarg_defaults
 from .. import show_layout
-from ...hdiplot import plot_hdi
 from ...kdeplot import _fast_kde
 
 
@@ -21,8 +21,8 @@ def plot_loo_pit(
     ecdf_fill,
     use_hdi,
     x_vals,
-    unif_densities,
     hdi_kwargs,
+    hdi_odds,
     n_unif,
     unif,
     plot_unif_kwargs,
@@ -43,7 +43,7 @@ def plot_loo_pit(
     if ax is None:
         backend_kwargs.setdefault("width", int(figsize[0] * dpi))
         backend_kwargs.setdefault("height", int(figsize[1] * dpi))
-        ax = bkp.figure(**backend_kwargs)
+        ax = bkp.figure(x_range=(0, 1), **backend_kwargs)
 
     if ecdf:
         if plot_kwargs.get("drawstyle") == "steps-mid":
@@ -115,20 +115,21 @@ def plot_loo_pit(
                 )
     else:
         if use_hdi:
-            plot_hdi(
-                x_vals,
-                unif_densities,
-                backend="bokeh",
-                ax=ax,
-                backend_kwargs={},
-                show=False,
-                **hdi_kwargs
+            ax.add_layout(
+                BoxAnnotation(
+                    bottom=hdi_odds[1],
+                    top=hdi_odds[0],
+                    fill_alpha=hdi_kwargs.pop("alpha"),
+                    fill_color=hdi_kwargs.pop("color"),
+                    **hdi_kwargs
+                )
             )
         else:
             for idx in range(n_unif):
-                unif_density, _, _ = _fast_kde(unif[idx, :], xmin=0, xmax=1)
+                unif_density, xmin, xmax = _fast_kde(unif[idx, :])
+                x_s = np.linspace(xmin, xmax, len(unif_density))
                 ax.line(
-                    x_vals,
+                    x_s,
                     unif_density,
                     line_color=plot_unif_kwargs.get("color", "black"),
                     line_alpha=plot_unif_kwargs.get("alpha", 0.1),

arviz/plots/backends/matplotlib/loopitplot.py

@@ -4,7 +4,6 @@
 
 from . import backend_kwarg_defaults, backend_show
 from ....numeric_utils import _fast_kde
-from ...hdiplot import plot_hdi
 
 
 def plot_loo_pit(
@@ -20,8 +19,8 @@ def plot_loo_pit(
     ecdf_fill,
     use_hdi,
     x_vals,
-    unif_densities,
     hdi_kwargs,
+    hdi_odds,
     n_unif,
     unif,
     plot_unif_kwargs,
@@ -54,14 +53,19 @@ def plot_loo_pit(
         else:
             ax.plot(unif_ecdf, p975 - unif_ecdf, unif_ecdf, p025 - unif_ecdf, **plot_unif_kwargs)
     else:
+        x_ss = np.empty((n_unif, len(loo_pit_kde)))
+        u_dens = np.empty((n_unif, len(loo_pit_kde)))
         if use_hdi:
-            plot_hdi(x_vals, unif_densities, **hdi_kwargs)
+            ax.axhspan(*hdi_odds, **hdi_kwargs)
         else:
             for idx in range(n_unif):
-                unif_density, _, _ = _fast_kde(unif[idx, :], xmin=0, xmax=1)
-                ax.plot(x_vals, unif_density, **plot_unif_kwargs)
+                unif_density, xmin, xmax = _fast_kde(unif[idx, :])
+                x_s = np.linspace(xmin, xmax, len(unif_density))
+                x_ss[idx] = x_s
+                u_dens[idx] = unif_density
+            ax.plot(x_ss.T, u_dens.T, **plot_unif_kwargs)
         ax.plot(x_vals, loo_pit_kde, **plot_kwargs)
-
+        ax.set_xlim(0, 1)
     ax.tick_params(labelsize=xt_labelsize)
     if legend:
         if not (use_hdi or (ecdf and ecdf_fill)):

arviz/plots/loopitplot.py

@@ -65,9 +65,9 @@ def plot_loo_pit(
     n_unif : int, optional
         Number of datasets to simulate and overlay from the uniform distribution.
     use_hdi : bool, optional
-        Use plot_hdi to fill between hdi values instead of overlaying the uniform distributions.
+        Compute expected hdi values instead of overlaying the sampled uniform distributions.
     credible_interval : float, optional
-        Credible interval of the hdi or of the ECDF theoretical credible interval
+        Theoretical credible interval. Works with ``use_hdi=True`` or ``ecdf=True``.
     figsize : figure size tuple, optional
         If None, size is (8 + numvars, 8 + numvars)
     textsize: int, optional
@@ -86,7 +86,7 @@ def plot_loo_pit(
         Additional keywords passed to ax.plot for overlaid uniform distributions or
         for beta credible interval lines if ``ecdf=True``
     hdi_kwargs : dict, optional
-        Additional keywords passed to az.plot_hdi
+        Additional keywords passed to ax.axhspan
     fill_kwargs : dict, optional
         Additional kwargs passed to ax.fill_between
     backend: str, optional
@@ -168,8 +168,8 @@ def plot_loo_pit(
     p975 = None
     p025 = None
     loo_pit_kde = None
+    hdi_odds = None
     unif = None
-    unif_densities = None
     x_vals = None
 
     if credible_interval is None:
@@ -206,20 +206,19 @@ def plot_loo_pit(
             )
             fill_kwargs.setdefault("label", "{:.3g}% credible interval".format(credible_interval))
     else:
-        loo_pit_kde, _, _ = _fast_kde(loo_pit, xmin=0, xmax=1)
+        loo_pit_kde, xmin, xmax = _fast_kde(loo_pit)
 
         unif = np.random.uniform(size=(n_unif, loo_pit.size))
-        x_vals = np.linspace(0, 1, len(loo_pit_kde))
+        x_vals = np.linspace(xmin, xmax, len(loo_pit_kde))
         if use_hdi:
+            n_obs = loo_pit.size
+            hdi_ = stats.beta(n_obs / 2, n_obs / 2).ppf((1 - credible_interval) / 2)
+            hdi_odds = (hdi_ / (1 - hdi_), (1 - hdi_) / hdi_)
             if hdi_kwargs is None:
                 hdi_kwargs = {}
             hdi_kwargs.setdefault("color", to_hex(hsv_to_rgb(light_color)))
-            hdi_fill_kwargs = hdi_kwargs.pop("fill_kwargs", {})
-            hdi_fill_kwargs.setdefault("label", "Uniform hdi")
-            hdi_kwargs["fill_kwargs"] = hdi_fill_kwargs
-            hdi_kwargs["credible_interval"] = credible_interval
-
-            unif_densities = np.empty((n_unif, len(loo_pit_kde)))
+            hdi_kwargs.setdefault("alpha", 0.35)
+            hdi_kwargs.setdefault("label", "Uniform hdi")
 
     loo_pit_kwargs = dict(
         ax=ax,
@@ -234,8 +233,8 @@ def plot_loo_pit(
         ecdf_fill=ecdf_fill,
         use_hdi=use_hdi,
         x_vals=x_vals,
-        unif_densities=unif_densities,
         hdi_kwargs=hdi_kwargs,
+        hdi_odds=hdi_odds,
         n_unif=n_unif,
         unif=unif,
         plot_unif_kwargs=plot_unif_kwargs,
@@ -253,14 +252,8 @@ def plot_loo_pit(
     backend = backend.lower()
 
     if backend == "bokeh":
-
-        if (
-            loo_pit_kwargs["hdi_kwargs"] is not None
-            and "fill_kwargs" in loo_pit_kwargs["hdi_kwargs"]
-            and loo_pit_kwargs["hdi_kwargs"]["fill_kwargs"] is not None
-            and "label" in loo_pit_kwargs["hdi_kwargs"]["fill_kwargs"]
-        ):
-            loo_pit_kwargs["hdi_kwargs"]["fill_kwargs"].pop("label")
+        if use_hdi:
+            loo_pit_kwargs["hdi_kwargs"].pop("label", None)
         loo_pit_kwargs.pop("legend")
         loo_pit_kwargs.pop("xt_labelsize")
         loo_pit_kwargs.pop("credible_interval")

arviz/tests/base_tests/test_plots_bokeh.py

@@ -603,8 +603,8 @@ def test_plot_khat_bad_input(models):
         {},
         {"n_unif": 50},
         {"use_hdi": True, "color": "gray"},
-        {"use_hdi": True, "credible_interval": 0.68, "plot_kwargs": {"alpha": 0.9}},
-        {"use_hdi": True, "hdi_kwargs": {"smooth": False}},
+        {"use_hdi": True, "credible_interval": 0.68},
+        {"use_hdi": True, "hdi_kwargs": {"line_dash": "dashed", "alpha": 0}},
         {"ecdf": True},
         {"ecdf": True, "ecdf_fill": False, "plot_unif_kwargs": {"line_dash": "--"}},
         {"ecdf": True, "credible_interval": 0.97, "fill_kwargs": {"color": "red"}},

arviz/tests/base_tests/test_plots_matplotlib.py

@@ -1101,8 +1101,8 @@ def test_plot_ess_no_divergences(models):
         {},
         {"n_unif": 50, "legend": False},
         {"use_hdi": True, "color": "gray"},
-        {"use_hdi": True, "credible_interval": 0.68, "plot_kwargs": {"ls": "--"}},
-        {"use_hdi": True, "hdi_kwargs": {"smooth": False}},
+        {"use_hdi": True, "credible_interval": 0.68},
+        {"use_hdi": True, "hdi_kwargs": {"fill": 0.1}},
         {"ecdf": True},
         {"ecdf": True, "ecdf_fill": False, "plot_unif_kwargs": {"ls": "--"}},
         {"ecdf": True, "credible_interval": 0.97, "fill_kwargs": {"hatch": "/"}},