UtilityTools
diff --git a/‎diff_diff/bootstrap_utils.py‎
Lines changed: 4 additions & 4 deletions b/‎diff_diff/bootstrap_utils.py‎
Lines changed: 4 additions & 4 deletions
diff --git a/‎diff_diff/prep.py‎
Lines changed: 6 additions & 0 deletions b/‎diff_diff/prep.py‎
Lines changed: 6 additions & 0 deletions
diff --git a/‎docs/methodology/REGISTRY.md‎
Lines changed: 3 additions & 2 deletions b/‎docs/methodology/REGISTRY.md‎
Lines changed: 3 additions & 2 deletions
diff --git a/‎tests/test_survey_phase8.py‎
Lines changed: 8 additions & 0 deletions b/‎tests/test_survey_phase8.py‎
Lines changed: 8 additions & 0 deletions
@@ -552,8 +552,8 @@ def generate_survey_multiplier_weights_batch(
                 pooled_cols = np.array(_singleton_cols)
                 weights[:, pooled_cols] = pooled_weights
             else:
-                # Single singleton — cannot pool, zero weight (matches R survey
-                # package behavior where adjust with one singleton = remove).
+                # Single singleton — cannot pool, zero weight (library-specific
+                # fallback; bootstrap adjust with one singleton = remove).
                 import warnings
 
                 warnings.warn(
@@ -684,8 +684,8 @@ def generate_rao_wu_weights(
                     p = int(obs_psu[idx])
                     rescaled[idx] = base_weights[idx] * psu_scale_map.get(p, 1.0)
         else:
-            # Single singleton — cannot pool, keep base weights (matches R
-            # survey package: adjust with one singleton = remove).
+            # Single singleton — cannot pool, keep base weights (library-specific
+            # fallback; bootstrap adjust with one singleton = remove).
             import warnings
 
             warnings.warn(
 
@@ -1280,6 +1280,12 @@ def trim_weights(
             raise ValueError(f"quantile must be in (0, 1), got {quantile}")
         upper = float(np.nanquantile(w, quantile))
 
+    if upper is not None and lower is not None and lower > upper:
+        raise ValueError(
+            f"lower ({lower}) must be <= upper ({upper}). "
+            f"When using quantile, the resolved upper cap may be below lower."
+        )
+
     if upper is not None:
         w = np.minimum(w, upper)
     if lower is not None:
 
@@ -2222,8 +2222,9 @@ ContinuousDiD, EfficientDiD):
   ContinuousDiD, EfficientDiD) and Rao-Wu bootstrap (SunAbraham, SyntheticDiD, TROP).
   FPC scaling is skipped for pooled singletons (conservative). When only one singleton
   stratum exists total, pooling is not possible — the singleton contributes zero bootstrap
-  variance (same as `remove`), with a `UserWarning` emitted. This matches R's `survey`
-  package behavior. Reference: Rust & Rao (1996).
+  variance (same as `remove`), with a `UserWarning` emitted. This is a library-specific
+  documented fallback (R's analytical `adjust` uses grand-mean centering, but the bootstrap
+  analogue for a single singleton is not defined in the literature). Reference: Rust & Rao (1996).
 - **Deviation from R:** For the no-FPC case (`m_h = n_h - 1`), this matches R
   `survey::as.svrepdesign(type="subbootstrap")`. The FPC-adjusted resample size
   `m_h = round((1-f_h)*(n_h-1))` follows Rao, Wu & Yue (1992) Section 3.
 
@@ -1201,6 +1201,14 @@ def test_trim_quantile_with_nan(self):
         assert np.isfinite(result["w"].iloc[0])
         assert np.isfinite(result["w"].iloc[4])
 
+    def test_trim_lower_exceeds_upper_raises(self):
+        """lower > upper raises ValueError."""
+        from diff_diff.prep import trim_weights
+
+        data = pd.DataFrame({"w": [1.0, 5.0, 10.0]})
+        with pytest.raises(ValueError, match="lower.*<=.*upper"):
+            trim_weights(data, "w", upper=5.0, lower=6.0)
+
 
 # ===========================================================================
 # 8e-iii: ImputationDiD Pretrends + Survey