Add verified DOIs to 11 bibliography entries

- 4 publisher DOIs via CrossRef: Spirtes (MIT Press), Hirano & Imbens
  (Wiley), Imbens & Rubin (Cambridge UP), Angrist & Pischke (Princeton UP)
- 6 arXiv DOIs (verified resolvable): NOTEARS, TARNet, DragonNet,
  DeepIV, DoWhy, CausalML
- Confirmed Angrist book DOI is 10.1515/9781400829828 (not the book
  review DOI suggested by editorialbot)
- 2 entries confirmed to have no DOI: Manski (1990, predates DOI system),
  EconML (software repo, no publication)

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

Author: brycewang-stanford

CHANGELOG.md

@@ -2,6 +2,31 @@
 
 All notable changes to StatsPAI will be documented in this file.
 
+## [0.6.2] - 2026-04-12
+
+### Added
+
+- **OLS `predict()`**: `result.predict(newdata=)` for out-of-sample prediction on OLS results
+- **`balance_panel()`**: Utility to keep only units observed in every time period (`sp.balance_panel()`)
+- **Panel `balance=True`**: Convenience flag in `sp.panel()` to auto-balance before estimation
+- **Analytical weights for DID**: `weights=` parameter added to `did()`, `ddd()`, and `event_study()` for population-weighted estimation (Stata `[aweight=...]` equivalent)
+- **Matching `ps_poly=`**: Polynomial propensity score specification (`ps_poly=2` adds interactions/squares, following Cunningham 2021 Ch. 5)
+- **Synth `rmspe` plot**: Post/pre RMSPE ratio histogram (`synthplot(result, type='rmspe')`) per Abadie et al. (2010)
+- **Synth placebo gap plot**: Full spaghetti placebo gap paths with `rmspe_threshold` filter (Abadie et al. 2010, Figure 4)
+- **Graddy (2006) replication**: Fulton Fish Market IV example added to `sp.replicate()` (Mixtape Ch. 7)
+- **Numerical validation tests**: Cross-validated against Stata/R reference values with humanized error messages
+
+### Fixed
+
+- **`outreg2` format auto-detection**: Correctly infers `.xlsx`/`.csv`/`.tex` from filename extension
+- **Synth placebo p-value**: Now uses RMSPE *ratio* (√post/√pre) instead of squared ratio, matching Abadie et al. (2010) convention
+
+### Improved
+
+- **DID/DDD/Event Study**: Weights propagation through WLS with proper normalization and validation
+- **Synth placebos**: Store full placebo gap trajectories, per-unit RMSPE ratios, and unit labels for richer post-estimation analysis
+- **Matching tests**: Added comprehensive test suite for PSM, Mahalanobis, CEM, and stratification methods
+
 ## [0.6.1] - 2026-04-07
 
 ### Fixed

README.md

@@ -558,6 +558,16 @@ Smart Workflow Engine (unique to StatsPAI):
 
 Interactive Plot Editor: Font presets redesigned to show actual font names; separate font and size presets for independent per-element control.
 
+### v0.6.2 (2026-04-12) — Weights, Prediction & Validation
+
+- **OLS `predict()`**: Out-of-sample prediction via `result.predict(newdata=)`
+- **`balance_panel()`**: Keep only units observed in every period
+- **DID/DDD/Event Study weights**: `weights=` parameter for population-weighted estimation
+- **Matching `ps_poly=`**: Polynomial propensity score models (Cunningham 2021, Ch. 5)
+- **Synth RMSPE plot**: `synthplot(result, type='rmspe')` histogram (Abadie et al. 2010)
+- **Graddy (2006) replication**: Fulton Fish Market IV example in `sp.replicate()`
+- **Numerical validation**: Cross-validated against Stata/R reference values
+
 ### v0.6.1 (2026-04-07) — Interactive Editor Fixes & Improvements
 
 - **Theme switching fix**: Themes now fully reset rcParams before applying, so switching between themes (e.g. ggplot → academic) correctly updates all visual properties

paper.bib

@@ -144,7 +144,8 @@ @inproceedings{zheng2018dags
   author={Zheng, Xun and Aragam, Bryon and Ravikumar, Pradeep K. and Xing, Eric P.},
   booktitle={Advances in Neural Information Processing Systems},
   volume={31},
-  year={2018}
+  year={2018},
+  doi={10.48550/arXiv.1803.01422}
 }
 
 @article{hahn2020bayesian,
@@ -188,23 +189,26 @@ @inproceedings{shalit2017estimating
   author={Shalit, Uri and Johansson, Fredrik D. and Sontag, David},
   booktitle={Proceedings of the 34th International Conference on Machine Learning},
   pages={3076--3085},
-  year={2017}
+  year={2017},
+  doi={10.48550/arXiv.1606.03976}
 }
 
 @inproceedings{shi2019adapting,
   title={Adapting neural networks for the estimation of treatment effects},
   author={Shi, Claudia and Blei, David and Veitch, Victor},
   booktitle={Advances in Neural Information Processing Systems},
   volume={32},
-  year={2019}
+  year={2019},
+  doi={10.48550/arXiv.1906.02120}
 }
 
 @book{spirtes2000causation,
   title={Causation, Prediction, and Search},
   author={Spirtes, Peter and Glymour, Clark and Scheines, Richard},
   year={2000},
   edition={2nd},
-  publisher={MIT Press}
+  publisher={MIT Press},
+  doi={10.7551/mitpress/1754.001.0001}
 }
 
 @article{lei2021conformal,
@@ -225,7 +229,8 @@ @incollection{hirano2004propensity
   booktitle={Applied Bayesian Modeling and Causal Inference from Incomplete-Data Perspectives},
   pages={73--84},
   year={2004},
-  publisher={Wiley}
+  publisher={Wiley},
+  doi={10.1002/0470090456.ch7}
 }
 
 @article{cattaneo2010efficient,
@@ -422,7 +427,8 @@ @inproceedings{hartford2017deep
   author={Hartford, Jason and Lewis, Greg and Leyton-Brown, Kevin and Taddy, Matt},
   booktitle={Proceedings of the 34th International Conference on Machine Learning},
   pages={1414--1423},
-  year={2017}
+  year={2017},
+  doi={10.48550/arXiv.1612.09596}
 }
 
 @article{kennedy2023towards,
@@ -454,22 +460,25 @@ @book{angrist2009mostly
   author={Angrist, Joshua D. and Pischke, J{\"o}rn-Steffen},
   year={2009},
   publisher={Princeton University Press},
-  isbn={978-0-691-12035-5}
+  isbn={978-0-691-12035-5},
+  doi={10.1515/9781400829828}
 }
 
 @book{imbens2015causal,
   title={Causal Inference for Statistics, Social, and Biomedical Sciences: An Introduction},
   author={Imbens, Guido W. and Rubin, Donald B.},
   year={2015},
   publisher={Cambridge University Press},
-  isbn={978-0-521-88588-1}
+  isbn={978-0-521-88588-1},
+  doi={10.1017/CBO9781139025751}
 }
 
 @article{sharma2020dowhy,
   title={{DoWhy}: An end-to-end library for causal inference},
   author={Sharma, Amit and Kiciman, Emre},
   journal={arXiv preprint arXiv:2011.04216},
-  year={2020}
+  year={2020},
+  doi={10.48550/arXiv.2011.04216}
 }
 
 @misc{econml,
@@ -483,5 +492,6 @@ @article{chen2020causalml
   title={{CausalML}: {Python} package for causal machine learning},
   author={Chen, Huigang and Harinen, Totte and Li, Jeong-Yoon and Yung, Liang and Zhao, Zhenyu},
   journal={arXiv preprint arXiv:2002.11631},
-  year={2020}
+  year={2020},
+  doi={10.48550/arXiv.2002.11631}
 }

pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "StatsPAI"
-version = "0.6.1"
+version = "0.6.2"
 description = "The Agent-Native Causal Inference & Econometrics Toolkit for Python"
 readme = "README.md"
 license = {text = "MIT"}

src/statspai/__init__.py

@@ -22,7 +22,7 @@
 >>> sp.outreg2(result, filename="results.xlsx")
 """
 
-__version__ = "0.6.1"
+__version__ = "0.6.2"
 __author__ = "Bryce Wang"
 __email__ = "bryce@copaper.ai"
 
@@ -55,7 +55,7 @@
 )
 from .dml import dml, DoubleML
 from .deepiv import deepiv, DeepIV
-from .panel import panel, panel_compare, PanelResults, PanelCompareResults, PanelRegression
+from .panel import panel, panel_compare, balance_panel, PanelResults, PanelCompareResults, PanelRegression
 from .causal_impact import causal_impact, CausalImpactEstimator, impactplot
 from .mediation import mediate, MediationAnalysis
 from .bartik import bartik, BartikIV, ssaggregate, shift_share_se
@@ -271,6 +271,7 @@
     # Panel
     "panel",
     "panel_compare",
+    "balance_panel",
     "PanelResults",
     "PanelCompareResults",
     "PanelRegression",

src/statspai/did/ddd.py

@@ -46,6 +46,7 @@ def ddd(
     cluster: Optional[str] = None,
     robust: bool = True,
     alpha: float = 0.05,
+    weights: Optional[str] = None,
 ) -> CausalResult:
     """
     Triple Differences (DDD) estimator.
@@ -77,6 +78,9 @@ def ddd(
         Use HC1 heteroskedasticity-robust standard errors.
     alpha : float, default 0.05
         Significance level for confidence intervals.
+    weights : str, optional
+        Column name for analytical weights (e.g. population weights).
+        Equivalent to Stata's ``[aweight=...]``.
 
     Returns
     -------
@@ -127,11 +131,14 @@ def ddd(
     # Triple interaction = DDD coefficient
     dtg = d * t * g
 
-    # Drop rows with NaN
+    # Drop rows with NaN (and invalid weights if provided)
     valid = np.isfinite(y_arr)
     if covariates:
         for cov in covariates:
             valid &= np.isfinite(df[cov].values.astype(float))
+    if weights is not None:
+        valid &= np.isfinite(df[weights].values.astype(float))
+        valid &= df[weights].values.astype(float) > 0
 
     d, t, g = d[valid], t[valid], g[valid]
     dt, dg, tg, dtg = dt[valid], dg[valid], tg[valid], dtg[valid]
@@ -155,13 +162,27 @@ def ddd(
     X = np.column_stack(X_parts)
     n, k = X.shape
 
-    # OLS
+    # --- Analytical weights (WLS) ---
+    if weights is not None:
+        w_raw = df.loc[valid, weights].values.astype(float) if isinstance(valid, np.ndarray) else df[weights].values.astype(float)
+        if np.any(w_raw < 0):
+            raise ValueError(f"Weights column '{weights}' contains negative values.")
+        w = w_raw * (n / w_raw.sum())
+        sqrt_w = np.sqrt(w)
+        Xw = X * sqrt_w[:, np.newaxis]
+        yw = y_arr * sqrt_w
+    else:
+        w = None
+        Xw = X
+        yw = y_arr
+
+    # OLS on (possibly weighted) data
     try:
-        XtX_inv = np.linalg.inv(X.T @ X)
+        XtX_inv = np.linalg.inv(Xw.T @ Xw)
     except np.linalg.LinAlgError:
-        XtX_inv = np.linalg.pinv(X.T @ X)
+        XtX_inv = np.linalg.pinv(Xw.T @ Xw)
 
-    beta = XtX_inv @ X.T @ y_arr
+    beta = XtX_inv @ Xw.T @ yw
     resid = y_arr - X @ beta
 
     # Variance-covariance
@@ -172,16 +193,25 @@ def ddd(
         meat = np.zeros((k, k))
         for c_val in unique_cl:
             idx = cl == c_val
-            score = (X[idx] * resid[idx, np.newaxis]).sum(axis=0)
+            if w is not None:
+                score = (Xw[idx] * (sqrt_w[idx] * resid[idx])[:, np.newaxis]).sum(axis=0)
+            else:
+                score = (X[idx] * resid[idx, np.newaxis]).sum(axis=0)
             meat += np.outer(score, score)
         correction = (n_cl / (n_cl - 1)) * ((n - 1) / (n - k))
         vcov = correction * XtX_inv @ meat @ XtX_inv
     elif robust:
-        weights = (n / (n - k)) * resid ** 2
-        meat = X.T @ np.diag(weights) @ X
+        if w is not None:
+            hc1_weights = (n / (n - k)) * (w * resid ** 2)
+        else:
+            hc1_weights = (n / (n - k)) * resid ** 2
+        meat = X.T @ np.diag(hc1_weights) @ X
         vcov = XtX_inv @ meat @ XtX_inv
     else:
-        sigma2 = np.sum(resid ** 2) / (n - k)
+        if w is not None:
+            sigma2 = np.sum(w * resid ** 2) / (n - k)
+        else:
+            sigma2 = np.sum(resid ** 2) / (n - k)
         vcov = sigma2 * XtX_inv
 
     se = np.sqrt(np.diag(vcov))
@@ -211,10 +241,15 @@ def ddd(
         'pvalue': pvals_all,
     })
 
-    # R-squared
-    tss = np.sum((y_arr - np.mean(y_arr)) ** 2)
-    rss = np.sum(resid ** 2)
-    r_squared = 1 - rss / tss
+    # R-squared (weighted if applicable)
+    if w is not None:
+        y_wmean = np.sum(w * y_arr) / np.sum(w)
+        tss = np.sum(w * (y_arr - y_wmean) ** 2)
+        rss = np.sum(w * resid ** 2)
+    else:
+        tss = np.sum((y_arr - np.mean(y_arr)) ** 2)
+        rss = np.sum(resid ** 2)
+    r_squared = 1 - rss / tss if tss > 0 else 0.0
 
     model_info = {
         'r_squared': round(r_squared, 6),
@@ -226,6 +261,7 @@ def ddd(
         'did_estimate': round(did_estimate, 6),
         'robust_se': robust,
         'cluster': cluster,
+        'weights': weights,
     }
 
     return CausalResult(