Add PrePostNEGD group-level summary outputs

Add experiment-specific summary statistics for PrePostNEGD by printing group-level sample sizes and pre/post means, and cover the new output in integration tests and the ANCOVA docs notebook.

Closes #727

Co-authored-by: Cursor <cursoragent@cursor.com>

Author: drbenvincent

.github/pr-summaries/727-add-prepostnegd-summary-stats.md

@@ -0,0 +1,30 @@
+# PR: Add group-level outputs to PrePostNEGD summary
+
+Closes #727
+
+## Issue Summary
+
+`PrePostNEGD.summary()` only reported formula, causal impact, and model coefficients. The issue requested experiment-specific outputs to make the summary more informative.
+
+## Root Cause
+
+The summary method had no implementation for experiment-specific diagnostics beyond the generic treatment-effect and coefficient output.
+
+## Solution
+
+Added a group-level descriptive statistics section to the `PrePostNEGD` summary output, including sample size and pre/post means by group, and added a regression test to verify those fields are printed.
+
+## Changes Made
+
+- `causalpy/experiments/prepostnegd.py`: added `_group_level_summary_stats()` and updated `summary()` to print group-level `n`, `pre_mean`, and `post_mean`.
+- `causalpy/tests/test_integration_pymc_examples.py`: updated `test_ancova` to capture summary output and assert new experiment-specific lines are present.
+
+## Testing
+
+- [x] Existing tests pass
+- [x] New tests added (if applicable)
+- [x] Manual verification completed
+
+## Notes
+
+The documentation notebook that displays this summary output and should be re-run is `docs/source/notebooks/ancova_pymc.ipynb`.

causalpy/experiments/prepostnegd.py

@@ -222,6 +222,23 @@ 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:
+            summary_df[["pre_mean", "post_mean"]] = summary_df[
+                ["pre_mean", "post_mean"]
+            ].round(round_to)
+        return summary_df
+
     def summary(self, round_to: int | None = None) -> None:
         """Print summary of main results and model coefficients.
 
@@ -230,6 +247,8 @@ 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 +576,11 @@ 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
     fig, ax = result.plot()
     assert isinstance(fig, plt.Figure)
     # For multi-panel plots, ax should be an array of axes