Address P2/P3: validate trim_weights caps, add survey pretrend coverage

P2: Validate upper/lower are finite and non-negative in trim_weights.
Reject NaN/negative/inf caps before applying np.minimum/np.maximum.

P3: Add survey+covariates pretrend test and survey+bootstrap pretrend
test exercising the distinct code paths.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

Author: igerber

diff_diff/prep.py

@@ -1280,6 +1280,13 @@ def trim_weights(
             raise ValueError(f"quantile must be in (0, 1), got {quantile}")
         upper = float(np.nanquantile(w, quantile))
 
+    # Validate cap values are finite and non-negative
+    if upper is not None:
+        if not np.isfinite(upper) or upper < 0:
+            raise ValueError(f"upper must be finite and >= 0, got {upper}")
+    if lower is not None:
+        if not np.isfinite(lower) or lower < 0:
+            raise ValueError(f"lower must be finite and >= 0, got {lower}")
     if upper is not None and lower is not None and lower > upper:
         raise ValueError(
             f"lower ({lower}) must be <= upper ({upper}). "

tests/test_survey_phase8.py

@@ -1283,6 +1283,49 @@ def test_pretrend_test_survey_no_raise(self, staggered_data):
             assert np.isfinite(pt["f_stat"])
             assert np.isfinite(pt["p_value"])
 
+    def test_pretrends_survey_with_covariates(self, staggered_data):
+        """Survey pretrends with covariates uses the survey+covariate path."""
+        from diff_diff import ImputationDiD
+
+        data = staggered_data
+        data["x1"] = np.random.default_rng(42).normal(0, 1, len(data))
+        sd = SurveyDesign(weights="weight", strata="stratum", psu="psu")
+        est = ImputationDiD(pretrends=True, horizon_max=3)
+        result = est.fit(
+            data,
+            outcome="outcome",
+            unit="unit",
+            time="time",
+            first_treat="first_treat",
+            covariates=["x1"],
+            survey_design=sd,
+            aggregate="event_study",
+        )
+        pre = [h for h in result.event_study_effects if h < -1]
+        for h in pre:
+            assert np.isfinite(result.event_study_effects[h]["se"])
+
+    def test_pretrends_survey_with_bootstrap(self, staggered_data):
+        """Survey pretrends + bootstrap: bootstrap doesn't overwrite pre-period SEs."""
+        from diff_diff import ImputationDiD
+
+        data = staggered_data
+        sd = SurveyDesign(weights="weight", strata="stratum", psu="psu")
+        est = ImputationDiD(pretrends=True, horizon_max=3, n_bootstrap=20, seed=42)
+        result = est.fit(
+            data,
+            outcome="outcome",
+            unit="unit",
+            time="time",
+            first_treat="first_treat",
+            survey_design=sd,
+            aggregate="event_study",
+        )
+        pre = [h for h in result.event_study_effects if h < -1]
+        for h in pre:
+            # Pre-period SEs come from lead regression, not bootstrap
+            assert np.isfinite(result.event_study_effects[h]["se"])
+
     def test_pretrends_survey_always_treated_psu(self):
         """Survey pretrends with a PSU/stratum that has no untreated obs."""
         from diff_diff import ImputationDiD