Wire allow_zero_se through survey bootstrap callers, add TROP deviation note

- Fix allow_zero_se to return p=NaN (not p=0.0) for census FPC
- Pass allow_zero_se=True from all 6 survey bootstrap callers (17 sites)
- Document TROP Rao-Wu frozen-tau as deviation (mathematically equivalent
  to refit since tau_{it} is deterministic given Y, D, lambda)
- Add TODO.md entry for survey bootstrap test coverage gaps

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

Author: igerber

TODO.md

@@ -78,6 +78,7 @@ Deferred items from PR reviews that were not addressed before merge.
 | CS R helpers hard-code `xformla = ~ 1`; no covariate-adjusted R benchmark for IRLS path | `tests/test_methodology_callaway.py` | #202 | Low |
 | ~376 `duplicate object description` Sphinx warnings — caused by autodoc `:members:` on dataclass attributes within manual API pages (not from autosummary stubs); fix requires restructuring `docs/api/*.rst` pages to avoid documenting the same attribute via both `:members:` and inline `autosummary` tables | `docs/api/*.rst` | — | Low |
 | Plotly renderers silently ignore styling kwargs (marker, markersize, linewidth, capsize, ci_linewidth) that the matplotlib backend honors; thread them through or reject when `backend="plotly"` | `visualization/_event_study.py`, `_diagnostic.py`, `_power.py` | #222 | Medium |
+| Survey bootstrap test coverage: add FPC census zero-variance, single-PSU NaN, full-design bootstrap for CS/ContinuousDiD/EfficientDiD, and TROP Rao-Wu vs block bootstrap equivalence tests | `tests/test_survey_phase*.py` | #237 | Medium |
 
 ---
 

diff_diff/bootstrap_utils.py

@@ -286,7 +286,7 @@ def compute_effect_bootstrap_stats(
     if not np.isfinite(se) or se <= 0:
         # Census FPC: all bootstrap estimates identical → SE=0 is legitimate
         if allow_zero_se and se == 0.0:
-            return 0.0, (original_effect, original_effect), 0.0
+            return 0.0, (original_effect, original_effect), np.nan
         warnings.warn(
             f"Bootstrap SE is non-finite or zero (n_valid={n_valid}) in {context}. "
             "Returning NaN for SE/CI/p-value.",
@@ -406,7 +406,7 @@ def compute_effect_bootstrap_stats_batch(
             ses[zero_idx] = 0.0
             ci_lowers[zero_idx] = original_effects[zero_idx]
             ci_uppers[zero_idx] = original_effects[zero_idx]
-            p_values[zero_idx] = 0.0
+            p_values[zero_idx] = np.nan  # p undefined when SE=0
         n_bad_se = int(np.sum(~se_valid & ~se_zero)) if allow_zero_se else int(np.sum(~se_valid))
         if n_bad_se > 0:
             warnings.warn(

diff_diff/continuous_did.py

@@ -1498,6 +1498,7 @@ def _bootstrap_gt_cell(gt, info):
                 boot_att_d[:, idx],
                 alpha=self.alpha,
                 context=f"ATT(d) at grid point {idx}",
+                allow_zero_se=_use_survey_bootstrap,
             )
             att_d_se[idx] = se
             att_d_ci_lower[idx] = ci[0]
@@ -1509,6 +1510,7 @@ def _bootstrap_gt_cell(gt, info):
                 boot_acrt_d[:, idx],
                 alpha=self.alpha,
                 context=f"ACRT(d) at grid point {idx}",
+                allow_zero_se=_use_survey_bootstrap,
             )
             acrt_d_se[idx] = se
             acrt_d_ci_lower[idx] = ci[0]
@@ -1530,6 +1532,7 @@ def _bootstrap_gt_cell(gt, info):
             boot_att_glob,
             alpha=self.alpha,
             context="overall ATT_glob",
+            allow_zero_se=_use_survey_bootstrap,
         )
         result["overall_att_se"] = se
         result["overall_att_ci"] = ci
@@ -1540,6 +1543,7 @@ def _bootstrap_gt_cell(gt, info):
             boot_acrt_glob,
             alpha=self.alpha,
             context="overall ACRT_glob",
+            allow_zero_se=_use_survey_bootstrap,
         )
         result["overall_acrt_se"] = se
         result["overall_acrt_ci"] = ci
@@ -1556,6 +1560,7 @@ def _bootstrap_gt_cell(gt, info):
                     boot_es[e],
                     alpha=self.alpha,
                     context=f"event study e={e}",
+                    allow_zero_se=_use_survey_bootstrap,
                 )
                 es_se[e] = se_e
                 es_ci[e] = ci_e

diff_diff/efficient_did_bootstrap.py

@@ -207,6 +207,7 @@ def _run_multiplier_bootstrap(
                 bootstrap_atts[:, j],
                 alpha=self.alpha,
                 context=f"ATT(g={gt[0]}, t={gt[1]})",
+                allow_zero_se=_use_survey_bootstrap,
             )
             gt_ses[gt] = se
             gt_cis[gt] = ci
@@ -220,6 +221,7 @@ def _run_multiplier_bootstrap(
                 bootstrap_overall,
                 alpha=self.alpha,
                 context="overall ATT",
+                allow_zero_se=_use_survey_bootstrap,
             )
 
         es_ses = es_cis = es_pvs = None
@@ -231,6 +233,7 @@ def _run_multiplier_bootstrap(
                     bootstrap_event_study[e],
                     alpha=self.alpha,
                     context=f"event study (e={e})",
+                    allow_zero_se=_use_survey_bootstrap,
                 )
                 es_ses[e] = se
                 es_cis[e] = ci
@@ -245,6 +248,7 @@ def _run_multiplier_bootstrap(
                     bootstrap_group[g],
                     alpha=self.alpha,
                     context=f"group effect (g={g})",
+                    allow_zero_se=_use_survey_bootstrap,
                 )
                 g_ses[g] = se
                 g_cis[g] = ci

diff_diff/imputation_bootstrap.py

@@ -349,6 +349,7 @@ def _run_bootstrap(
             boot_overall_shifted,
             alpha=self.alpha,
             context="ImputationDiD overall ATT",
+            allow_zero_se=_use_survey_bootstrap,
         )
 
         event_study_ses = None
@@ -366,6 +367,7 @@ def _run_bootstrap(
                     shifted_h,
                     alpha=self.alpha,
                     context=f"ImputationDiD event study (h={h})",
+                    allow_zero_se=_use_survey_bootstrap,
                 )
                 event_study_ses[h] = se_h
                 event_study_cis[h] = ci_h
@@ -386,6 +388,7 @@ def _run_bootstrap(
                     shifted_g,
                     alpha=self.alpha,
                     context=f"ImputationDiD group effect (g={g})",
+                    allow_zero_se=_use_survey_bootstrap,
                 )
                 group_ses[g] = se_g
                 group_cis[g] = ci_g

diff_diff/staggered_bootstrap.py

@@ -426,7 +426,8 @@ def _run_multiplier_bootstrap(
 
         # Batch compute bootstrap statistics for ATT(g,t)
         batch_ses, batch_ci_lo, batch_ci_hi, batch_pv = _compute_effect_bootstrap_stats_batch_func(
-            original_atts, bootstrap_atts_gt, alpha=self.alpha
+            original_atts, bootstrap_atts_gt, alpha=self.alpha,
+            allow_zero_se=_use_survey_bootstrap,
         )
         gt_ses = {}
         gt_cis = {}
@@ -442,8 +443,9 @@ def _run_multiplier_bootstrap(
             overall_ci = (np.nan, np.nan)
             overall_p_value = np.nan
         else:
-            overall_se, overall_ci, overall_p_value = self._compute_effect_bootstrap_stats(
-                original_overall, bootstrap_overall, context="overall ATT"
+            overall_se, overall_ci, overall_p_value = _compute_effect_bootstrap_stats_func(
+                original_overall, bootstrap_overall, alpha=self.alpha, context="overall ATT",
+                allow_zero_se=_use_survey_bootstrap,
             )
 
         # Batch compute bootstrap statistics for event study effects
@@ -455,7 +457,8 @@ def _run_multiplier_bootstrap(
             es_effects = np.array([event_study_info[e]["effect"] for e in rel_periods])
             es_boot_matrix = np.column_stack([bootstrap_event_study[e] for e in rel_periods])
             es_ses, es_ci_lo, es_ci_hi, es_pv = _compute_effect_bootstrap_stats_batch_func(
-                es_effects, es_boot_matrix, alpha=self.alpha
+                es_effects, es_boot_matrix, alpha=self.alpha,
+                allow_zero_se=_use_survey_bootstrap,
             )
             event_study_ses = {e: float(es_ses[i]) for i, e in enumerate(rel_periods)}
             event_study_cis = {
@@ -472,7 +475,8 @@ def _run_multiplier_bootstrap(
             grp_effects = np.array([group_agg_info[g]["effect"] for g in group_list])
             grp_boot_matrix = np.column_stack([bootstrap_group[g] for g in group_list])
             grp_ses, grp_ci_lo, grp_ci_hi, grp_pv = _compute_effect_bootstrap_stats_batch_func(
-                grp_effects, grp_boot_matrix, alpha=self.alpha
+                grp_effects, grp_boot_matrix, alpha=self.alpha,
+                allow_zero_se=_use_survey_bootstrap,
             )
             group_effect_ses = {g: float(grp_ses[i]) for i, g in enumerate(group_list)}
             group_effect_cis = {

diff_diff/sun_abraham.py

@@ -1476,6 +1476,7 @@ def _run_rao_wu_bootstrap(
                 boot_dist,
                 alpha=self.alpha,
                 context=f"event study e={e}",
+                allow_zero_se=True,
             )
             event_study_ses[e] = se
             event_study_cis[e] = ci
@@ -1487,6 +1488,7 @@ def _run_rao_wu_bootstrap(
             bootstrap_overall,
             alpha=self.alpha,
             context="overall ATT",
+            allow_zero_se=True,
         )
 
         return SABootstrapResults(

diff_diff/two_stage_bootstrap.py

@@ -311,6 +311,7 @@ def _run_bootstrap(
             boot_overall_shifted,
             alpha=self.alpha,
             context="TwoStageDiD overall ATT",
+            allow_zero_se=_use_survey_bootstrap,
         )
 
         # --- Event study bootstrap ---
@@ -411,6 +412,7 @@ def _run_bootstrap(
                         shifted_h,
                         alpha=self.alpha,
                         context=f"TwoStageDiD event study (h={h})",
+                        allow_zero_se=_use_survey_bootstrap,
                     )
                     event_study_ses[h] = se_h
                     event_study_cis[h] = ci_h
@@ -474,6 +476,7 @@ def _run_bootstrap(
                     shifted_g,
                     alpha=self.alpha,
                     context=f"TwoStageDiD group effect (g={g})",
+                    allow_zero_se=_use_survey_bootstrap,
                 )
                 group_ses[g] = se_g
                 group_cis[g] = ci_g

docs/methodology/REGISTRY.md

@@ -1986,6 +1986,11 @@ ContinuousDiD, EfficientDiD):
   TROP uses treatment group (treated vs control) as pseudo-strata for Rao-Wu resampling
   to preserve treatment ratio. FPC is applied within these pseudo-strata. This matches
   TROP's existing treatment-stratified resampling pattern.
+- **Note (deviation from block bootstrap):** In Rao-Wu survey bootstrap, per-observation
+  treatment effects tau_{it} are deterministic given (Y, D, lambda) because survey weights
+  do not enter the kernel-weighted matrix completion. The Rao-Wu path therefore precomputes
+  tau values once and only varies the ATT aggregation weights across draws. This is
+  mathematically equivalent to refitting per draw and avoids redundant computation.
 
 ---