plot_cap() can have different model parameters as output (#627)

* plot_cap gains a new argument 'target'

* Add seed to test and update changelog

* Update tests/test_plots.py

Co-authored-by: Ravin Kumar <7213793+canyon289@users.noreply.github.com>

* Update tests/test_plots.py

Co-authored-by: Ravin Kumar <7213793+canyon289@users.noreply.github.com>

* Update tests/test_plots.py

Co-authored-by: Ravin Kumar <7213793+canyon289@users.noreply.github.com>

* remove old 'print'

* fix indentation

Co-authored-by: Ravin Kumar <7213793+canyon289@users.noreply.github.com>

Author: tomicapretto

Changelog.md

@@ -6,6 +6,7 @@
 
 * Refactored the codebase to support distributional models (#607)
 * Added a default method to handle posterior predictive sampling for custom families (#625)
+* `plot_cap()` gains a new argument `target` that allows to plot different parameters of the response distribution (#627)
 
 ### Maintenance and fixes
 

bambi/plots/plot_cap.py

@@ -128,6 +128,7 @@ def plot_cap(
     model,
     idata,
     covariates,
+    target="mean",
     use_hdi=True,
     hdi_prob=None,
     transforms=None,
@@ -152,6 +153,8 @@ def plot_cap(
         and the third is mapped to different plot panels.
         If a dictionary, keys must be taken from ("horizontal", "color", "panel") and the values
         are the names of the variables.
+    target : str
+        Which model parameter to plot. Defaults to 'mean'.
     use_hdi : bool, optional
         Whether to compute the highest density interval (defaults to True) or the quantiles.
     hdi_prob : float, optional
@@ -203,11 +206,11 @@ def plot_cap(
     response_name = get_aliased_name(model.response_component.response_term)
     response_transform = transforms.get(response_name, identity)
 
-    y_hat = response_transform(idata.posterior[f"{response_name}_mean"])
+    y_hat = response_transform(idata.posterior[f"{response_name}_{target}"])
     y_hat_mean = y_hat.mean(("chain", "draw"))
 
     if use_hdi:
-        y_hat_bounds = az.hdi(y_hat, hdi_prob)[f"{response_name}_mean"].T
+        y_hat_bounds = az.hdi(y_hat, hdi_prob)[f"{response_name}_{target}"].T
     else:
         lower_bound = round((1 - hdi_prob) / 2, 4)
         upper_bound = 1 - lower_bound
@@ -222,7 +225,6 @@ def plot_cap(
         axes = np.atleast_1d(axes)
     else:
         axes = np.atleast_1d(ax)
-        print(axes)
         if isinstance(axes[0], np.ndarray):
             fig = axes[0][0].get_figure()
         else:
@@ -238,8 +240,9 @@ def plot_cap(
     else:
         raise ValueError("Main covariate must be numeric or categoric.")
 
+    ylabel = response_name if target == "mean" else target
     for ax in axes.ravel():  # pylint: disable = redefined-argument-from-local
-        ax.set(xlabel=main, ylabel=response_name)
+        ax.set(xlabel=main, ylabel=ylabel)
 
     return fig, axes
 

tests/test_plots.py

@@ -5,7 +5,7 @@
 import matplotlib.pyplot as plt
 import pytest
 
-from bambi.models import Model
+from bambi.models import Model, Formula
 from bambi.plots import plot_cap
 
 
@@ -150,3 +150,23 @@ def test_transforms(mtcars):
 
     transforms = {"mpg": np.log, "hp": np.log}
     plot_cap(model, idata, ["hp"], transforms=transforms)
+
+
+def test_multiple_outputs():
+    """Test plot cap default and specified values for target argument"""
+    rng = np.random.default_rng(121195)
+    N = 200
+    a, b = 0.5, 1.1
+    x = rng.uniform(-1.5, 1.5, N)
+    shape = np.exp(0.3 + x * 0.5 + rng.normal(scale=0.1, size=N))
+    y = rng.gamma(shape, np.exp(a + b * x) / shape, N)
+    data_gamma = pd.DataFrame({"x": x, "y": y})
+
+
+    formula = Formula("y ~ x", "alpha ~ x")
+    model = Model(formula, data_gamma, family="gamma")
+    idata = model.fit(tune=100, draws=100, random_seed=1234)
+    # Test default target 
+    plot_cap(model, idata, "x")
+    # Test user supplied target argument
+    plot_cap(model, idata, "x", "alpha")