fix: plot seasonalities of logistic models with minmax scaling and no floor

python/prophet/plot.py

@@ -309,7 +309,11 @@ def seasonality_plot_df(
     -------
     A dataframe with seasonal components on ds.
     """
-    df_dict = {'ds': ds, 'cap': 1., 'floor': 0.}
+    floor = 0.
+    if not m.logistic_floor and m.scaling == 'minmax':
+        # setup_dataframe sets the floor to y_min, so keep the cap above it.
+        floor = cast(float, m.y_min)
+    df_dict = {'ds': ds, 'cap': floor + 1., 'floor': floor}
     for name in m.extra_regressors:
         df_dict[name] = 0.
     # Activate all conditional seasonality columns

python/prophet/tests/test_plot.py

@@ -162,6 +162,19 @@ def test_without_uncertainty_samples(self, daily_univariate_ts, backend):
         data = _seasonality_plot_data(m, "weekly", days, uncertainty=True)
         assert list(data.columns) == ["ds", "weekly"]
 
+    def test_logistic_minmax_without_floor(self, daily_univariate_ts, backend):
+        # With minmax scaling and no floor column, the floor is the minimum of y.
+        history = daily_univariate_ts.copy()
+        history["cap"] = history["y"].max() * 1.5
+        assert history["y"].min() > 1
+        m = Prophet(stan_backend=backend, growth="logistic", scaling="minmax")
+        m.fit(history, seed=1237861298)
+        days = pd.date_range("2017-01-01", periods=7)
+        data = _seasonality_plot_data(m, "weekly", days)
+        fcst = m.predict(pd.DataFrame({"ds": days, "cap": history["cap"].iloc[0]}))
+        for col in ["weekly", "weekly_lower", "weekly_upper"]:
+            np.testing.assert_allclose(data[col], fcst[col])
+
     @pytest.mark.parametrize("uncertainty", [True, False])
     def test_matplotlib_draws_data(self, model_and_forecast, uncertainty):
         m, _ = model_and_forecast