Get plot data for prepostfit experiments

causalpy/experiments/base.py

@@ -17,6 +17,7 @@
 
 from abc import abstractmethod
 
+import pandas as pd
 from sklearn.base import RegressorMixin
 
 from causalpy.pymc_models import PyMCModel
@@ -59,22 +60,45 @@
     def plot(self, *args, **kwargs) -> tuple:
         """Plot the model.
 
-        Internally, this function dispatches to either `bayesian_plot` or `ols_plot`
+        Internally, this function dispatches to either `_bayesian_plot` or `_ols_plot`
         depending on the model type.
         """
         if isinstance(self.model, PyMCModel):
-            return self.bayesian_plot(*args, **kwargs)
+            return self._bayesian_plot(*args, **kwargs)
         elif isinstance(self.model, RegressorMixin):
-            return self.ols_plot(*args, **kwargs)
+            return self._ols_plot(*args, **kwargs)
         else:
             raise ValueError("Unsupported model type")
 
     @abstractmethod
-    def bayesian_plot(self, *args, **kwargs):
+    def _bayesian_plot(self, *args, **kwargs):
         """Abstract method for plotting the model."""
-        raise NotImplementedError("bayesian_plot method not yet implemented")
+        raise NotImplementedError("_bayesian_plot method not yet implemented")
 
     @abstractmethod
-    def ols_plot(self, *args, **kwargs):
+    def _ols_plot(self, *args, **kwargs):
         """Abstract method for plotting the model."""
-        raise NotImplementedError("ols_plot method not yet implemented")
+        raise NotImplementedError("_ols_plot method not yet implemented")
+
+    def get_plot_data(self, *args, **kwargs) -> pd.DataFrame:
+        """Recover the data of a PrePostFit experiment along with the prediction and causal impact information.
+
+        Internally, this function dispatches to either :func:`get_plot_data_bayesian` or :func:`get_plot_data_ols`
+        depending on the model type.
+        """
+        if isinstance(self.model, PyMCModel):
+            return self.get_plot_data_bayesian(*args, **kwargs)
+        elif isinstance(self.model, RegressorMixin):
+            return self.get_plot_data_ols(*args, **kwargs)
+        else:
+            raise ValueError("Unsupported model type")
+
+    @abstractmethod
+    def get_plot_data_bayesian(self, *args, **kwargs):
+        """Abstract method for recovering plot data."""
+        raise NotImplementedError("get_plot_data_bayesian method not yet implemented")
+
+    @abstractmethod
+    def get_plot_data_ols(self, *args, **kwargs):
+        """Abstract method for recovering plot data."""
+        raise NotImplementedError("get_plot_data_ols method not yet implemented")

causalpy/experiments/diff_in_diff.py

@@ -229,7 +229,7 @@ def _causal_impact_summary_stat(self, round_to=None) -> str:
         """Computes the mean and 94% credible interval bounds for the causal impact."""
         return f"Causal impact = {convert_to_string(self.causal_impact, round_to=round_to)}"
 
-    def bayesian_plot(self, round_to=None, **kwargs) -> tuple[plt.Figure, plt.Axes]:
+    def _bayesian_plot(self, round_to=None, **kwargs) -> tuple[plt.Figure, plt.Axes]:
         """
         Plot the results
 
@@ -367,7 +367,7 @@ def _plot_causal_impact_arrow(results, ax):
         )
         return fig, ax
 
-    def ols_plot(self, round_to=None, **kwargs) -> tuple[plt.Figure, plt.Axes]:
+    def _ols_plot(self, round_to=None, **kwargs) -> tuple[plt.Figure, plt.Axes]:
         """Generate plot for difference-in-differences"""
         round_to = kwargs.get("round_to")
         fig, ax = plt.subplots()

causalpy/experiments/prepostfit.py

@@ -25,7 +25,7 @@
 from sklearn.base import RegressorMixin
 
 from causalpy.custom_exceptions import BadIndexException
-from causalpy.plot_utils import plot_xY
+from causalpy.plot_utils import get_hdi_to_df, plot_xY
 from causalpy.pymc_models import PyMCModel
 from causalpy.utils import round_num
 
@@ -123,7 +123,7 @@
         print(f"Formula: {self.formula}")
         self.print_coefficients(round_to)
 
-    def bayesian_plot(
+    def _bayesian_plot(
         self, round_to=None, **kwargs
     ) -> tuple[plt.Figure, List[plt.Axes]]:
         """
@@ -231,7 +231,7 @@
 
         return fig, ax
 
-    def ols_plot(self, round_to=None, **kwargs) -> tuple[plt.Figure, List[plt.Axes]]:
+    def _ols_plot(self, round_to=None, **kwargs) -> tuple[plt.Figure, List[plt.Axes]]:
         """
         Plot the results
 
@@ -303,6 +303,70 @@
 
         return (fig, ax)
 
+    def get_plot_data_bayesian(self, hdi_prob: float = 0.94) -> pd.DataFrame:
+        """
+        Recover the data of a PrePostFit experiment along with the prediction and causal impact information.
+
+        :param hdi_prob:
+            Prob for which the highest density interval will be computed. The default value is defined as the default from the :func:`arviz.hdi` function.
+        """
+        if isinstance(self.model, PyMCModel):
+            hdi_pct = int(round(hdi_prob * 100))
+
+            pred_lower_col = f"pred_hdi_lower_{hdi_pct}"
+            pred_upper_col = f"pred_hdi_upper_{hdi_pct}"
+            impact_lower_col = f"impact_hdi_lower_{hdi_pct}"
+            impact_upper_col = f"impact_hdi_upper_{hdi_pct}"
+
+            pre_data = self.datapre.copy()
+            post_data = self.datapost.copy()
+
+            pre_data["prediction"] = (
+                az.extract(self.pre_pred, group="posterior_predictive", var_names="mu")
+                .mean("sample")
+                .values
+            )
+            post_data["prediction"] = (
+                az.extract(self.post_pred, group="posterior_predictive", var_names="mu")
+                .mean("sample")
+                .values
+            )
+            pre_data[[pred_lower_col, pred_upper_col]] = get_hdi_to_df(
+                self.pre_pred["posterior_predictive"].mu, hdi_prob=hdi_prob
+            ).set_index(pre_data.index)
+            post_data[[pred_lower_col, pred_upper_col]] = get_hdi_to_df(
+                self.post_pred["posterior_predictive"].mu, hdi_prob=hdi_prob
+            ).set_index(post_data.index)
+
+            pre_data["impact"] = self.pre_impact.mean(dim=["chain", "draw"]).values
+            post_data["impact"] = self.post_impact.mean(dim=["chain", "draw"]).values
+            pre_data[[impact_lower_col, impact_upper_col]] = get_hdi_to_df(
+                self.pre_impact, hdi_prob=hdi_prob
+            ).set_index(pre_data.index)
+            post_data[[impact_lower_col, impact_upper_col]] = get_hdi_to_df(
+                self.post_impact, hdi_prob=hdi_prob
+            ).set_index(post_data.index)
+
+            self.plot_data = pd.concat([pre_data, post_data])
+
+            return self.plot_data
+        else:
+            raise ValueError("Unsupported model type")
+
+    def get_plot_data_ols(self) -> pd.DataFrame:
+        """
+        Recover the data of a PrePostFit experiment along with the prediction and causal impact information.
+        """
+        pre_data = self.datapre.copy()
+        post_data = self.datapost.copy()
+        pre_data["prediction"] = self.pre_pred
+        post_data["prediction"] = self.post_pred
+        pre_data["impact"] = self.pre_impact
+        post_data["impact"] = self.post_impact
+        self.plot_data = pd.concat([pre_data, post_data])
+
+        return self.plot_data
+
 
 class InterruptedTimeSeries(PrePostFit):
     """
@@ -382,7 +446,7 @@
     supports_ols = True
     supports_bayes = True
 
-    def bayesian_plot(self, *args, **kwargs) -> tuple[plt.Figure, List[plt.Axes]]:
+    def _bayesian_plot(self, *args, **kwargs) -> tuple[plt.Figure, List[plt.Axes]]:
         """
         Plot the results
 
@@ -393,7 +457,7 @@
             Whether to plot the control units as well. Defaults to False.
         """
         # call the super class method
-        fig, ax = super().bayesian_plot(*args, **kwargs)
+        fig, ax = super()._bayesian_plot(*args, **kwargs)
 
         # additional plotting functionality for the synthetic control experiment
         plot_predictors = kwargs.get("plot_predictors", False)

causalpy/experiments/prepostnegd.py

@@ -200,7 +200,7 @@ def summary(self, round_to=None) -> None:
         print(self._causal_impact_summary_stat(round_to))
         self.print_coefficients(round_to)
 
-    def bayesian_plot(
+    def _bayesian_plot(
         self, round_to=None, **kwargs
     ) -> tuple[plt.Figure, List[plt.Axes]]:
         """Generate plot for ANOVA-like experiments with non-equivalent group designs."""

causalpy/experiments/regression_discontinuity.py

@@ -218,7 +218,7 @@ def summary(self, round_to=None) -> None:
         print("\n")
         self.print_coefficients(round_to)
 
-    def bayesian_plot(self, round_to=None, **kwargs) -> tuple[plt.Figure, plt.Axes]:
+    def _bayesian_plot(self, round_to=None, **kwargs) -> tuple[plt.Figure, plt.Axes]:
         """Generate plot for regression discontinuity designs."""
         fig, ax = plt.subplots()
         # Plot raw data
@@ -267,7 +267,7 @@ def bayesian_plot(self, round_to=None, **kwargs) -> tuple[plt.Figure, plt.Axes]:
         )
         return (fig, ax)
 
-    def ols_plot(self, round_to=None, **kwargs) -> tuple[plt.Figure, plt.Axes]:
+    def _ols_plot(self, round_to=None, **kwargs) -> tuple[plt.Figure, plt.Axes]:
         """Generate plot for regression discontinuity designs."""
         fig, ax = plt.subplots()
         # Plot raw data

causalpy/experiments/regression_kink.py

@@ -189,7 +189,7 @@ def summary(self, round_to=None) -> None:
         )
         self.print_coefficients(round_to)
 
-    def bayesian_plot(self, round_to=None, **kwargs) -> tuple[plt.Figure, plt.Axes]:
+    def _bayesian_plot(self, round_to=None, **kwargs) -> tuple[plt.Figure, plt.Axes]:
         """Generate plot for regression kink designs."""
         fig, ax = plt.subplots()
         # Plot raw data

causalpy/plot_utils.py

@@ -79,3 +79,24 @@ def plot_xY(
         filter(lambda x: isinstance(x, PolyCollection), ax_hdi.get_children())
     )[-1]
     return (h_line, h_patch)
+
+
+def get_hdi_to_df(
+    x: xr.DataArray,
+    hdi_prob: float = 0.94,
+) -> pd.DataFrame:
+    """
+    Utility function to calculate and recover HDI intervals.
+
+    :param x:
+        Xarray data array
+    :param hdi_prob:
+        The size of the HDI, default is 0.94
+    """
+    hdi = (
+        az.hdi(x, hdi_prob=hdi_prob)
+        .to_dataframe()
+        .unstack(level="hdi")
+        .droplevel(0, axis=1)
+    )
+    return hdi