add predictive_finder method (experimental)

preliz/internal/parser.py

@@ -54,7 +54,10 @@ def parse_arguments(lst, regex):
     return result
 
 
-def get_prior_pp_samples(fmodel, draws):
+def get_prior_pp_samples(fmodel, draws, values=None):
+    if values is None:
+        values = []
+
     match = match_return_variables(fmodel)
     if match:
         variables = [var.strip() for var in match.group(1).split(",")]
@@ -63,7 +66,7 @@ def get_prior_pp_samples(fmodel, draws):
     pp_samples_ = []
     prior_samples_ = {name: [] for name in variables[:-1]}
     for _ in range(draws):
-        for name, value in zip(variables, fmodel()):
+        for name, value in zip(variables, fmodel(*values)):
             if name == obs_rv:
                 pp_samples_.append(value)
             else:

preliz/internal/plot_helper.py

@@ -405,47 +405,51 @@ def looper(*args, **kwargs):
         _, ax = plt.subplots()
         ax.set_xlim(x_min, x_max, auto=auto)
 
-        alpha = max(0.01, 1 - iterations * 0.009)
-
-        if kind_plot == "hist":
-            if results[0].dtype.kind == "i":
-                bins = np.arange(np.min(results), np.max(results) + 1.5) - 0.5
-                if len(bins) < 30:
-                    ax.set_xticks(bins + 0.5)
-            else:
-                bins = "auto"
-            ax.hist(
-                results,
-                alpha=alpha,
-                density=True,
-                color=["C0"] * iterations,
-                bins=bins,
-                histtype="step",
-            )
-            ax.hist(
-                np.concatenate(results),
-                density=True,
-                bins=bins,
-                color="k",
-                ls="--",
-                histtype="step",
-            )
-        elif kind_plot == "kde":
-            for result in results:
-                plt.plot(*_kde_linear(result, grid_len=100), "C0", alpha=alpha)
-            plt.plot(*_kde_linear(np.concatenate(results), grid_len=100), "k--")
-        elif kind_plot == "ecdf":
-            plt.plot(
-                np.sort(results, axis=1).T,
-                np.linspace(0, 1, len(results[0]), endpoint=False),
-                color="C0",
-            )
-            a = np.concatenate(results)
-            plt.plot(np.sort(a), np.linspace(0, 1, len(a), endpoint=False), "k--")
+        plot_repr(results, kind_plot, iterations, ax)
 
     return looper
 
 
+def plot_repr(results, kind_plot, iterations, ax):
+    alpha = max(0.01, 1 - iterations * 0.009)
+
+    if kind_plot == "hist":
+        if results[0].dtype.kind == "i":
+            bins = np.arange(np.min(results), np.max(results) + 1.5) - 0.5
+            if len(bins) < 30:
+                ax.set_xticks(bins + 0.5)
+        else:
+            bins = "auto"
+        ax.hist(
+            results,
+            alpha=alpha,
+            density=True,
+            color=["0.5"] * iterations,
+            bins=bins,
+            histtype="step",
+        )
+        ax.hist(
+            np.concatenate(results),
+            density=True,
+            bins=bins,
+            color="k",
+            ls="--",
+            histtype="step",
+        )
+    elif kind_plot == "kde":
+        for result in results:
+            ax.plot(*_kde_linear(result, grid_len=100), "0.5", alpha=alpha)
+        ax.plot(*_kde_linear(np.concatenate(results), grid_len=100), "k--")
+    elif kind_plot == "ecdf":
+        ax.plot(
+            np.sort(results, axis=1).T,
+            np.linspace(0, 1, len(results[0]), endpoint=False),
+            color="0.5",
+        )
+        a = np.concatenate(results)
+        ax.plot(np.sort(a), np.linspace(0, 1, len(a), endpoint=False), "k--")
+
+
 def plot_pp_samples(pp_samples, pp_samples_idxs, references, kind="pdf", sharex=True, fig=None):
     row_colum = int(np.ceil(len(pp_samples_idxs) ** 0.5))
 
@@ -576,3 +580,27 @@ def representations(fitted_dist, kind_plot, ax):
     elif kind_plot == "ppf":
         fitted_dist.plot_ppf(pointinterval=True, legend="title", ax=ax)
         ax.set_xlim(-0.01, 1)
+
+
+def create_figure(figsize):
+    """
+    Initialize a matplotlib figure with one subplot
+    """
+    fig, axes = plt.subplots(1, 1, figsize=figsize, constrained_layout=True)
+    axes.set_yticks([])
+    fig.canvas.header_visible = False
+    fig.canvas.footer_visible = False
+    fig.canvas.toolbar_position = "right"
+
+    return fig, axes
+
+
+def reset_dist_panel(ax, yticks):
+    """
+    Clean the distribution subplot
+    """
+    ax.cla()
+    if yticks:
+        ax.set_yticks([])
+    ax.relim()
+    ax.autoscale_view()

preliz/internal/predictive_helper.py

@@ -0,0 +1,70 @@
+from sys import modules
+
+import numpy as np
+
+
+from .plot_helper import (
+    repr_to_matplotlib,
+)
+
+from ..unidimensional import mle
+
+
+def back_fitting(model, subset, new_families=True):
+    """
+    Use MLE to fit a subset of the prior samples to the marginal prior distributions
+    """
+    string = "Your selection is consistent with the priors (original families):\n"
+
+    for name, dist in model.items():
+        dist._fit_mle(subset[name])
+        string += f"{name} = {repr_to_matplotlib(dist)}\n"
+
+    if new_families:
+        string += "\nYour selection is consistent with the priors (new families):\n"
+
+        exclude, distributions = get_distributions()
+        for name, dist in model.items():
+            if dist.__class__.__name__ in exclude:
+                dist._fit_mle(subset[name])
+            else:
+                idx, _ = mle(distributions, subset[name], plot=False)
+                dist = distributions[idx[0]]
+            string += f"{name} = {repr_to_matplotlib(dist)}\n"
+
+    return string, np.concatenate([dist.params for dist in model.values()])
+
+
+def get_distributions():
+    exclude = [
+        "Beta",
+        "BetaScaled",
+        "Triangular",
+        "TruncatedNormal",
+        "Uniform",
+        "VonMises",
+        "Categorical",
+        "DiscreteUniform",
+        "HyperGeometric",
+        "zeroInflatedBinomial",
+        "ZeroInflatedNegativeBinomial",
+        "ZeroInflatedPoisson",
+        "MvNormal",
+    ]
+    all_distributions = modules["preliz.distributions"].__all__
+    distributions = []
+    for a_dist in all_distributions:
+        dist = getattr(modules["preliz.distributions"], a_dist)()
+        if dist.__class__.__name__ not in exclude:
+            distributions.append(dist)
+    return exclude, distributions
+
+
+def select_prior_samples(selected, prior_samples, model):
+    """
+    Given a selected set of prior predictive samples pick the corresponding
+    prior samples.
+    """
+    subsample = {rv: prior_samples[rv][selected] for rv in model.keys()}
+
+    return subsample

preliz/predictive/__init__.py

@@ -1,5 +1,6 @@
 from .ppa import ppa
 from .predictive_explorer import predictive_explorer
+from .predictive_finder import predictive_finder
 
 
-__all__ = ["ppa", "predictive_explorer"]
+__all__ = ["ppa", "predictive_explorer", "predictive_finder"]

preliz/predictive/ppa.py

@@ -2,7 +2,6 @@
 
 import logging
 from random import shuffle
-from sys import modules
 
 try:
     import ipywidgets as widgets
@@ -17,12 +16,11 @@
     check_inside_notebook,
     plot_pp_samples,
     plot_pp_mean,
-    repr_to_matplotlib,
 )
 from ..internal.parser import inspect_source, parse_function_for_ppa, get_prior_pp_samples
+from ..internal.predictive_helper import back_fitting, select_prior_samples
 from ..distributions.continuous import Normal
 from ..distributions.distributions import Distribution
-from ..unidimensional import mle
 
 _log = logging.getLogger("preliz")
 
@@ -33,12 +31,11 @@ def ppa(
     """
     Prior predictive check assistant.
 
-    This is experimental
+    This is an experimental method under development, use with caution.
 
     Parameters
     ----------
     model : PreliZ model
-        Model associated to ``idata``.
     draws : int
         Number of draws from the prior and prior predictive distribution
     summary : str
@@ -176,7 +173,7 @@ def on_return_prior(fig, selected, model):
             if len(selected) > 4:
                 subsample = select_prior_samples(selected, prior_samples, model)
 
-                string = back_fitting(model, subsample)
+                string, _ = back_fitting(model, subsample)
 
                 fig.clf()
                 plt.text(0.05, 0.5, string, fontsize=14)
@@ -412,63 +409,3 @@ def collect_more_samples(
         return selected, shown
     else:
         return selected, shown
-
-
-def select_prior_samples(selected, prior_samples, model):
-    """
-    Given a selected set of prior predictive samples pick the corresponding
-    prior samples.
-    """
-    subsample = {rv: prior_samples[rv][selected] for rv in model.keys()}
-
-    return subsample
-
-
-def back_fitting(model, subset, new_families=True):
-    """
-    Use MLE to fit a subset of the prior samples to the marginal prior distributions
-    """
-    string = "Your selection is consistent with the priors (original families):\n"
-
-    for name, dist in model.items():
-        dist._fit_mle(subset[name])
-        string += f"{name} = {repr_to_matplotlib(dist)}\n"
-
-    if new_families:
-        string += "\nYour selection is consistent with the priors (new families):\n"
-
-        exclude, distributions = get_distributions()
-        for name, dist in model.items():
-            if dist.__class__.__name__ in exclude:
-                dist._fit_mle(subset[name])
-            else:
-                idx, _ = mle(distributions, subset[name], plot=False)
-                dist = distributions[idx[0]]
-            string += f"{name} = {repr_to_matplotlib(dist)}\n"
-
-    return string
-
-
-def get_distributions():
-    exclude = [
-        "Beta",
-        "BetaScaled",
-        "Triangular",
-        "TruncatedNormal",
-        "Uniform",
-        "VonMises",
-        "Categorical",
-        "DiscreteUniform",
-        "HyperGeometric",
-        "zeroInflatedBinomial",
-        "ZeroInflatedNegativeBinomial",
-        "ZeroInflatedPoisson",
-        "MvNormal",
-    ]
-    all_distributions = modules["preliz.distributions"].__all__
-    distributions = []
-    for a_dist in all_distributions:
-        dist = getattr(modules["preliz.distributions"], a_dist)()
-        if dist.__class__.__name__ not in exclude:
-            distributions.append(dist)
-    return exclude, distributions