Add group-level outputs to PrePostNEGD.summary()

.gitignore

@@ -20,10 +20,13 @@ docs/source/api/generated/
 
 # Temporary files generated by Cursor commands
 .github/pr_summaries/
+.github/pr-summaries/
 
 # Issue evaluation drafts generated by evaluate_issue command
 .github/issue_comments/
+.github/issue-comments/
 .github/issue_summaries/
+.github/issue-summaries/
 
 # Local scratch space (not tracked)
 .scratch/

causalpy/experiments/prepostnegd.py

@@ -222,6 +222,22 @@ def _causal_impact_summary_stat(self, round_to: int | None = 2) -> str:
         causal_impact = f"{round_num(self.causal_impact.mean(), round_to)}, "
         return f"Causal impact = {causal_impact + ci}"
 
+    def _group_level_summary_stats(self, round_to: int | None = None) -> pd.DataFrame:
+        """Compute group-level sample sizes and pre/post means."""
+        summary_df = (
+            self.data.groupby(self.group_variable_name)
+            .agg(
+                n=(self.group_variable_name, "size"),
+                pre_mean=(self.pretreatment_variable_name, "mean"),
+                post_mean=(self.outcome_variable_name, "mean"),
+            )
+            .reset_index()
+        )
+        if round_to is not None:
+            for col in ["pre_mean", "post_mean"]:
+                summary_df[col] = summary_df[col].map(lambda x: round_num(x, round_to))
+        return summary_df
+
     def summary(self, round_to: int | None = None) -> None:
         """Print summary of main results and model coefficients.
 
@@ -230,6 +246,8 @@ def summary(self, round_to: int | None = None) -> None:
         """
         print(f"{self.expt_type:=^80}")
         print(f"Formula: {self.formula}")
+        print("\nGroup-level descriptive stats:")
+        print(self._group_level_summary_stats(round_to).to_string(index=False))
         print("\nResults:")
         print(self._causal_impact_summary_stat(round_to))
         self.print_coefficients(round_to)

causalpy/tests/test_integration_pymc_examples.py

@@ -553,7 +553,7 @@ def test_sc_brexit(mock_pymc_sample):
 
 
 @pytest.mark.integration
-def test_ancova(mock_pymc_sample):
+def test_ancova(mock_pymc_sample, capsys):
     """
     Test Pre-PostNEGD experiment on anova1 data.
 
@@ -576,6 +576,17 @@ def test_ancova(mock_pymc_sample):
     assert len(result.idata.posterior.coords["chain"]) == sample_kwargs["chains"]
     assert len(result.idata.posterior.coords["draw"]) == sample_kwargs["draws"]
     result.summary()
+    output = capsys.readouterr().out
+    assert "Group-level descriptive stats:" in output
+    assert "n" in output
+    assert "pre_mean" in output
+    assert "post_mean" in output
+    rounded_stats = result._group_level_summary_stats(round_to=1)
+    assert isinstance(rounded_stats, pd.DataFrame)
+    for col in ["pre_mean", "post_mean"]:
+        assert pd.api.types.is_string_dtype(rounded_stats[col])
+        for val in rounded_stats[col]:
+            float(val)
     fig, ax = result.plot()
     assert isinstance(fig, plt.Figure)
     # For multi-panel plots, ax should be an array of axes