Fix Step 3 example column and align evaluation doc labels

P1: Complete example Step 3 now uses treatment_group='treat' (binary
ever-treated indicator) instead of 'first_treat' (cohort year), with
explicit pre_periods=[2003, 2004, 2005]. The previous version would
produce an empty treated subset since check_parallel_trends() splits
on == 1 / == 0.

P2: Evaluation doc comparison tables and narrative now use canonical
step labels (S3=Test PT, S4=Choose estimator, S5=Estimate) matching
REGISTRY.md and the practitioner guide.

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

Author: igerber

docs/llms-practitioner.txt

@@ -390,8 +390,11 @@ data = load_mpdta()
 #          no anticipation, doubly robust for conditional PT
 
 # Step 3: Test parallel trends
+# Note: check_parallel_trends expects a binary treatment indicator, not cohort years.
+# Use 'treat' (ever-treated binary) with pre-periods before any cohort adopts.
 pt = check_parallel_trends(data, outcome='lemp', time='year',
-                           treatment_group='first_treat')
+                           treatment_group='treat',
+                           pre_periods=[2003, 2004, 2005])
 print(f"Pre-trends p-value: {pt['p_value']:.4f}")
 
 # Step 4: Choose estimator — staggered adoption → CS (primary), SA (robustness)

docs/practitioner-guide-evaluation.md

@@ -51,9 +51,9 @@ instances with no shared context.
 |------|------------|-----------|--------|----------------|
 | S1: Target parameter | 1.0 | 2.0 | **+1.0** | Agents now explicitly define weighted/unweighted target |
 | S2: Assumptions | 1.0 | 2.0 | **+1.0** | Agents now formally name PT variant (PT-GT-NYT) |
-| S3: Estimator choice | 2.0 | 2.0 | 0.0 | Already perfect before |
-| S4: Inference | 1.0 | 1.5 | +0.5 | Now discuss wild bootstrap alternative |
-| S5: Estimation | 2.0 | 2.0 | 0.0 | Already perfect before |
+| S3: Test parallel trends | 0.1 | 2.0 | **+1.9** | From near-zero to universal formal PT testing |
+| S4: Choose estimator | 2.0 | 2.0 | 0.0 | Already perfect before |
+| S5: Estimate (cluster check) | 1.0 | 1.5 | +0.5 | Now discuss wild bootstrap alternative |
 | S6: Sensitivity | **0.1** | **2.0** | **+1.9** | From near-zero to universal HonestDiD + placebo |
 | S7: Heterogeneity | 1.4 | 2.0 | +0.6 | Now consistently do group + event study |
 | S8: Robustness | 0.9 | 1.75 | +0.85 | Now compare 3 estimators; ~50% add with/without covariates |
@@ -77,9 +77,9 @@ instances with no shared context.
    standardized behavior — agents now consistently follow the same high-quality
    workflow rather than producing variable-quality ad hoc analyses.
 
-5. **Steps 3 and 5 were already perfect**: Agents already knew to use CS for
-   staggered data and could produce working code. The gap was never in mechanics
-   but in empiricist reasoning.
+5. **Steps 4 and 5 (estimator choice + estimation) were already perfect**:
+   Agents already knew to use CS for staggered data and could produce working
+   code. The gap was never in mechanics but in empiricist reasoning.
 
 ## Qualitative Observations
 
@@ -101,8 +101,8 @@ instances with no shared context.
 ## Iteration 2: Targeted Fixes
 
 After v1, the remaining 0.75 point gap was concentrated in:
-- **Step 4 (Inference)**: Agents mentioned wild bootstrap generically but never
-  checked the actual cluster count in the data (1.5/2 across all runs).
+- **Step 5 (Estimate/Inference)**: Agents mentioned wild bootstrap generically but
+  never checked the actual cluster count in the data (1.5/2 across all runs).
 - **Step 8 (Robustness)**: ~50% of agents skipped with/without covariates
   comparison despite the guide listing it (mean 1.75/2).
 
@@ -129,9 +129,9 @@ After v1, the remaining 0.75 point gap was concentrated in:
 |------|--------|----------|----------|
 | S1: Target parameter | 1.0 | 2.0 | 2.0 |
 | S2: Assumptions | 1.0 | 2.0 | 2.0 |
-| S3: Estimator choice | 2.0 | 2.0 | 2.0 |
-| S4: Inference | 1.0 | 1.5 | **2.0** |
-| S5: Estimation | 2.0 | 2.0 | 2.0 |
+| S3: Test parallel trends | 0.1 | 2.0 | 2.0 |
+| S4: Choose estimator | 2.0 | 2.0 | 2.0 |
+| S5: Estimate (cluster check) | 1.0 | 1.5 | **2.0** |
 | S6: Sensitivity | 0.1 | 2.0 | 2.0 |
 | S7: Heterogeneity | 1.4 | 2.0 | 2.0 |
 | S8: Robustness | 0.9 | 1.75 | **2.0** |