arxiv: 2605.07834 · v1 · submitted 2026-05-08 · 📊 stat.ME · stat.AP

Recognition: 2 theorem links

· Lean Theorem

GenAI Powered Dynamic Causal Inference with Unstructured Data

Kentaro Nakamura , Kosuke Imai

Authors on Pith no claims yet

Pith reviewed 2026-05-11 02:21 UTC · model grok-4.3

classification 📊 stat.ME stat.AP

keywords causal inferencegenerative AIunstructured datadynamic treatmentsdeconfoundersmarginal structural modelsneural networkssequence effects

0 comments

The pith

A generative AI framework enables valid causal inference on sequences of treatment features within text, images, and video.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper develops a statistical framework for dynamic causal inference with unstructured data by first extracting internal representations from generative AI models. It then feeds these into a neural network architecture that estimates a marginal structural model while jointly learning a deconfounder for each treatment feature and its position in the sequence. This setup produces asymptotic confidence intervals for the causal effects of ordered treatments. Previous approaches treated each object as a single static unit and could not capture how placement or order within the data affects outcomes. Simulations show the estimator recovers target effects with proper coverage, and an application to a Hong Kong protests experiment finds that a feature's causal impact depends on its position in the text.

Core claim

By extracting internal representations from a GenAI model and using a neural network architecture to jointly learn a deconfounder for each treatment feature in the sequence, the method estimates a marginal structural model that yields valid asymptotic confidence intervals for the causal effects of sequences of treatment features in unstructured data.

What carries the argument

Neural network architecture that jointly learns a deconfounder for each treatment feature in the sequence, using internal representations extracted from a generative AI model.

If this is right

Causal effects of treatment features can be estimated while accounting for their specific positions within sequences of text or video.
Asymptotic confidence intervals remain valid for these dynamic causal effects under the stated conditions.
In finite samples the estimator recovers the target causal effects and the intervals achieve nominal coverage.
The effect of a given treatment feature depends on its position in the sequence, as demonstrated when the method is applied to randomized protest messaging data.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

The same approach could be tested on video data to measure how the timing of visual or spoken features causally shapes viewer responses.
It opens analysis of causal order effects in variable-length social media posts where message composition varies across users.
Researchers might combine the extracted representations with other sequence models to handle mixed text-image inputs in a single causal framework.

Load-bearing premise

The internal representations extracted from the GenAI model together with the neural network architecture are sufficient to learn a valid deconfounder for each treatment feature in the sequence without residual confounding or model misspecification.

What would settle it

In repeated simulation studies with known target causal effects, the estimator failing to recover those effects or the confidence intervals failing to achieve nominal coverage in finite samples would show the framework does not deliver valid inference.

Figures

Figures reproduced from arXiv: 2605.07834 by Kentaro Nakamura, Kosuke Imai.

**Figure 2.** Figure 2: Diagram Illustrating the Proposed Model Architecture. The proposed model takes an [PITH_FULL_IMAGE:figures/full_fig_p010_2.png] view at source ↗

**Figure 3.** Figure 3: Average estimated potential outcomes (blue) versus their oracle values (black) across [PITH_FULL_IMAGE:figures/full_fig_p016_3.png] view at source ↗

**Figure 4.** Figure 4: Performance of the estimator across different sample sizes and values of the incremental [PITH_FULL_IMAGE:figures/full_fig_p017_4.png] view at source ↗

**Figure 5.** Figure 5: Estimated average potential outcomes under stochastic interventions that perturb the [PITH_FULL_IMAGE:figures/full_fig_p018_5.png] view at source ↗

read the original abstract

A growing number of scholars seek to estimate causal effects of unstructured data such as text, images, and video. However, existing methods typically treat each object as a single, static observation. We develop a statistical framework for dynamic causal inference with unstructured data by leveraging generative artificial intelligence (GenAI) models. Our approach enables researchers to estimate the causal effects of sequences of treatment features, including their positions within text and video. We first extract internal representations of unstructured objects from a GenAI model and then estimate a marginal structural model using a neural network architecture that jointly learns a deconfounder for each treatment feature in the sequence. Our semiparametric inference framework yields valid asymptotic confidence intervals. Simulation studies demonstrate that the proposed estimator recovers the target causal effects and that the confidence intervals achieve nominal coverage in finite samples. We further apply our method to a randomized experiment on the Hong Kong protests, showing that the effect of a treatment feature depends critically on its position within the text.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

The paper gives a framework for position-dependent causal effects in text or video sequences via GenAI embeddings and a neural deconfounder, but the asymptotic CI claim rests on unverified assumptions that those components fully eliminate confounding.

read the letter

The main point is a method to estimate causal effects of ordered features inside unstructured data like text, pulling internal representations from a GenAI model and feeding them into a neural network that jointly deconfounds each position inside a marginal structural model. This moves past the usual static-document setup and lets the effect of a feature depend on where it sits in the sequence. The Hong Kong protests application shows the idea in a randomized experiment, and the simulations are reported to recover the target effects with nominal coverage for the intervals. That combination of GenAI representations with the NN deconfounder for dynamic unstructured data is not in the prior static-text causal work they cite, so the integration itself is new. The soft spot is the identification step. The asymptotic normality and valid confidence intervals require that the pretrained GenAI embeddings plus the network architecture span all relevant confounders for every position without residual bias or approximation error. In dynamic settings, time-varying confounders can depend on earlier treatments and outcomes, and general-purpose GenAI models trained on unrelated data have no guarantee of capturing those. The abstract gives no details on the simulation data-generating processes, hyperparameter choices, or sensitivity checks, so it is hard to judge how fragile the coverage results are. This is for social scientists and causal-inference researchers who work with text, images, or video and need tools that respect sequence and position. A reader looking for a concrete proposal to handle dynamic unstructured treatments would find it worth reading, though they would have to verify the deconfounding assumptions in their own data. It deserves a serious referee because it targets a genuine gap with a workable estimator and an empirical example. I would recommend sending it to peer review.

Referee Report

2 major / 2 minor

Summary. The paper develops a statistical framework for dynamic causal inference with unstructured data (text, images, video) by leveraging generative AI models. It extracts internal representations from GenAI models and uses a neural network architecture to jointly learn deconfounders for estimating marginal structural models on sequences of treatment features. The framework provides semiparametric inference for asymptotic confidence intervals, supported by simulation studies showing effect recovery and nominal coverage, and an application to a randomized experiment on Hong Kong protests demonstrating position-dependent effects.

Significance. If the core assumptions hold—namely that GenAI representations combined with the NN architecture can learn valid deconfounders without residual confounding—this work could significantly advance causal inference methods for dynamic, unstructured data settings, which are increasingly common in social sciences. The semiparametric approach and real-data application are strengths, though the reliance on learned representations introduces challenges not fully addressed in standard theory.

major comments (2)

Semiparametric inference framework (as described in the abstract and methods): The claim that the framework yields valid asymptotic confidence intervals depends on the GenAI internal representations and NN jointly learning a complete deconfounder without residual confounding or approximation error; however, no explicit rate conditions on the NN approximation error or completeness of the embedding space are provided, which is critical for the dynamic treatment setting where time-varying confounders may not be fully captured by general-purpose GenAI representations.
Simulation studies: The reported recovery of target causal effects and nominal coverage in finite samples lacks accompanying details on data-generating processes, hyperparameter choices for the NN, or sensitivity checks, making it difficult to verify robustness to the modeling assumptions underlying the central claims.

minor comments (2)

Abstract: The description of how the neural network jointly learns a deconfounder for each treatment feature in the sequence could be expanded for clarity on the architecture and loss function.
Application: Specify the exact GenAI model used and the preprocessing steps for extracting treatment features from the Hong Kong protests text data.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their constructive and insightful comments, which have helped us identify areas to strengthen the manuscript. We address each major comment below and indicate the revisions we plan to make.

read point-by-point responses

Referee: Semiparametric inference framework (as described in the abstract and methods): The claim that the framework yields valid asymptotic confidence intervals depends on the GenAI internal representations and NN jointly learning a complete deconfounder without residual confounding or approximation error; however, no explicit rate conditions on the NN approximation error or completeness of the embedding space are provided, which is critical for the dynamic treatment setting where time-varying confounders may not be fully captured by general-purpose GenAI representations.

Authors: We appreciate the referee's emphasis on the conditions required for valid asymptotic inference in this setting. The framework assumes that the GenAI-derived representations, when processed through the neural network architecture, capture the necessary deconfounders for the marginal structural model without residual confounding. In the revised manuscript, we will add an explicit statement of this assumption, including a discussion of the completeness of the embedding space for time-varying confounders in dynamic treatment regimes. While providing explicit convergence rates for general-purpose pretrained GenAI models is challenging and outside the primary scope of the work, we will clarify that the semiparametric results hold under the condition that any approximation error vanishes at an appropriate rate relative to the sample size, drawing parallels to existing semiparametric causal inference literature that employs machine learning components. This addition will better contextualize the theoretical guarantees. revision: partial
Referee: Simulation studies: The reported recovery of target causal effects and nominal coverage in finite samples lacks accompanying details on data-generating processes, hyperparameter choices for the NN, or sensitivity checks, making it difficult to verify robustness to the modeling assumptions underlying the central claims.

Authors: We agree that greater detail on the simulation design is necessary to allow readers to assess robustness and reproducibility. In the revised manuscript, we will expand the simulation studies section (and associated appendix) to provide full specifications of the data-generating processes, including how unstructured data sequences and confounders are simulated; complete details on neural network hyperparameters such as architecture, layer dimensions, activation functions, regularization, and optimization settings; and results from sensitivity analyses that vary key parameters and modeling choices. These revisions will enable direct verification of the reported effect recovery and coverage properties. revision: yes

Circularity Check

0 steps flagged

No significant circularity in the derivation chain.

full rationale

The paper proposes a new semiparametric framework that extracts internal representations from pretrained GenAI models and employs a neural network to jointly learn deconfounders for sequential treatment features in unstructured data, then applies this to marginal structural models for causal effect estimation. The claim of valid asymptotic confidence intervals follows from standard semiparametric theory once the nuisance functions (deconfounders) are estimated at appropriate rates, without reducing the target estimand to any fitted parameter or input by construction. No self-definitional steps, fitted-input predictions, or load-bearing self-citations appear; the validity hinges on external assumptions about representation completeness and approximation quality, which the paper tests via simulations rather than assuming tautologically. The approach is self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

1 free parameters · 2 axioms · 1 invented entities

The central claim rests on standard causal assumptions plus the untested claim that GenAI representations plus a neural network suffice to deconfound dynamic treatments. No new physical entities are postulated.

free parameters (1)

neural network architecture and hyperparameters
The joint learning of deconfounders for each treatment feature requires choosing network depth, width, regularization, and optimization settings that are fitted to data.

axioms (2)

domain assumption No unmeasured confounding conditional on the learned representations and observed covariates
Invoked when the neural network is said to learn a valid deconfounder for the marginal structural model.
standard math Positivity and consistency assumptions of the marginal structural model hold for the sequence of treatment features
Required for identification of the dynamic causal effects.

invented entities (1)

learned deconfounder for each treatment feature no independent evidence
purpose: To remove confounding for position-dependent effects in the sequence
The neural network is trained to produce this deconfounder; no independent evidence outside the model fit is provided.

pith-pipeline@v0.9.0 · 5459 in / 1466 out tokens · 32469 ms · 2026-05-11T02:21:00.834368+00:00 · methodology

discussion (0)

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

IndisputableMonolith/Cost/FunctionalEquation.lean washburn_uniqueness_aczel unclear

?

unclear
Relation between the paper passage and the cited Recognition theorem.

We first extract internal representations of unstructured objects from a GenAI model and then estimate a marginal structural model using a neural network architecture that jointly learns a deconfounder for each treatment feature in the sequence. Our semiparametric inference framework yields valid asymptotic confidence intervals.
IndisputableMonolith/Foundation/RealityFromDistinction.lean reality_from_one_distinction unclear

?

unclear
Relation between the paper passage and the cited Recognition theorem.

Theorem 3 (Asymptotic Normality) ... √N(Ψ̂(δ)−Ψ(δ))/σ(δ) →d G(δ)

What do these tags mean?

matches: The paper's claim is directly supported by a theorem in the formal canon.
supports: The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends: The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses: The paper appears to rely on the theorem as machinery.
contradicts: The paper's claim conflicts with a theorem or certificate in the canon.
unclear: Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

300 extracted references · 300 canonical work pages · 2 internal anchors

[1]

American Political Science Review , volume=

How large and long-lasting are the persuasive effects of televised campaign ads? Results from a randomized field experiment , author=. American Political Science Review , volume=. 2011 , publisher=

work page 2011
[2]

Sociological Methods & Research , volume=

First equals most important? Order effects in vignette-based measurement , author=. Sociological Methods & Research , volume=. 2017 , publisher=

work page 2017
[3]

arXiv preprint arXiv:2410.14812 , year=

Isolated causal effects of natural language , author=. arXiv preprint arXiv:2410.14812 , year=

work page arXiv
[4]

arXiv preprint arXiv:2602.15730 , year=

Causal Effect Estimation with Latent Textual Treatments , author=. arXiv preprint arXiv:2602.15730 , year=

work page arXiv
[5]

Estimating causal effects of text interventions leveraging LLMs , author=. Proceedings of the 14th International Joint Conference on Natural Language Processing and the 4th Conference of the Asia-Pacific Chapter of the Association for Computational Linguistics , pages=

work page
[6]

Text and Code Embeddings by Contrastive Pre-Training

Text and code embeddings by contrastive pre-training , author=. arXiv preprint arXiv:2201.10005 , year=

work page internal anchor Pith review arXiv
[7]

Proceedings of the 47th International ACM SIGIR Conference on Research and Development in Information Retrieval , pages=

Fine-tuning llama for multi-stage text retrieval , author=. Proceedings of the 47th International ACM SIGIR Conference on Research and Development in Information Retrieval , pages=

work page
[8]

Scaling sentence embeddings with large language models.arXiv preprint arXiv:2307.16645, 2023

Scaling sentence embeddings with large language models , author=. arXiv preprint arXiv:2307.16645 , year=

work page arXiv
[9]

Early Childhood Research Quarterly , volume=

Text to talk: Effects of a home-school vocabulary texting intervention on prekindergarten vocabulary , author=. Early Childhood Research Quarterly , volume=. 2022 , publisher=

work page 2022
[10]

NPJ Digital Medicine , volume=

Randomized controlled study using text messages to help connect new medicaid beneficiaries to primary care , author=. NPJ Digital Medicine , volume=. 2021 , publisher=

work page 2021
[11]

and Wellner, Jon A

van der Vaart, Aad W. and Wellner, Jon A. , title =. 1996 , isbn =

work page 1996
[12]

Public opinion quarterly , volume=

An evaluation of a cognitive theory of response-order effects in survey measurement , author=. Public opinion quarterly , volume=. 1987 , publisher=

work page 1987
[13]

arXiv preprint arXiv:2507.03897 , year=

Genai-powered inference , author=. arXiv preprint arXiv:2507.03897 , year=

work page arXiv
[14]

Cosmos World Foundation Model Platform for Physical AI

Cosmos world foundation model platform for physical ai , author=. arXiv preprint arXiv:2501.03575 , year=

work page internal anchor Pith review Pith/arXiv arXiv
[15]

American Political Science Review , volume=

The effect of television advertising in United States elections , author=. American Political Science Review , volume=. 2022 , publisher=

work page 2022
[16]

Electoral Studies , volume=

Candidate appearance in campaign advertisements , author=. Electoral Studies , volume=. 2021 , publisher=

work page 2021
[17]

American Political Science Review , volume=

The effect of TV ads and candidate appearances on statewide presidential votes, 1988--96 , author=. American Political Science Review , volume=. 1999 , publisher=

work page 1988
[18]

Journal of the Royal Statistical Society, Series B (Statistical Methodology) , year =

Papadogeorgou, Georgia and Imia, Kosuke and Lyall, Jason and Li, Fan , title =. Journal of the Royal Statistical Society, Series B (Statistical Methodology) , year =

work page
[19]

Journal of the American Statistical Association , year =

Imai, Kosuke and Jiang, Zhichao , title =. Journal of the American Statistical Association , year =

work page
[20]

2018 , publisher=

Improving language understanding by generative pre-training , author=. 2018 , publisher=

work page 2018
[21]

arXiv preprint arXiv:1705.08582 , year=

On the multiply robust estimation of the mean of the g-functional , author=. arXiv preprint arXiv:1705.08582 , year=

work page arXiv
[22]

Biometrics , volume=

Population intervention causal effects based on stochastic interventions , author=. Biometrics , volume=. 2012 , publisher=

work page 2012
[23]

Working Paper , year=

Replication for Language Models Problems, Principles, and Best Practice for Political Science , author=. Working Paper , year=

work page
[24]

2024 , eprint=

Gemma: Open Models Based on Gemini Research and Technology , author=. 2024 , eprint=

work page 2024
[25]

Journal of Business & Economic Statistics , number=

Double debiased machine learning nonparametric inference with continuous treatments , author=. Journal of Business & Economic Statistics , number=. 2025 , publisher=

work page 2025
[26]

American Political Science Review , year =

Egami, Naoki and Hartman, Erin , title =. American Political Science Review , year =

work page
[27]

2024 , url =

Llama 3 Model Card , author=. 2024 , url =

work page 2024
[28]

J., Ting, D

Thirunavukarasu, Arun James and Ting, Darren Shu Jeng and Elangovan, Kabilan and Gutierrez, Laura and Tan, Ting Fang and Ting, Daniel Shu Wei , date =. Large language models in medicine , url =. Nature Medicine , number =. 2023 , bdsk-url-1 =. doi:10.1038/s41591-023-02448-8 , id =

work page doi:10.1038/s41591-023-02448-8 2023
[29]

Learning and Individual Differences

Enkelejda Kasneci and Kathrin Sessler and Stefan Küchemann and Maria Bannert and Daryna Dementieva and Frank Fischer and Urs Gasser and Georg Groh and Stephan Günnemann and Eyke Hüllermeier and Stephan Krusche and Gitta Kutyniok and Tilman Michaeli and Claudia Nerdel and Jürgen Pfeffer and Oleksandra Poquet and Michael Sailer and Albrecht Schmidt and Tina...

work page doi:10.1016/j.lindif.2023.102274 2023
[30]

Liu , title =

Colin Raffel and Noam Shazeer and Adam Roberts and Katherine Lee and Sharan Narang and Michael Matena and Yanqi Zhou and Wei Li and Peter J. Liu , title =. Journal of Machine Learning Research , year =

work page
[31]

Dwivedi and Thomas H

Nir Kshetri and Yogesh K. Dwivedi and Thomas H. Davenport and Niki Panteli , keywords =. Generative artificial intelligence in marketing: Applications, opportunities, challenges, and research agenda , journal =. 2024 , issn =. doi:https://doi.org/10.1016/j.ijinfomgt.2023.102716 , url =

work page doi:10.1016/j.ijinfomgt.2023.102716 2024
[32]

Political Analysis , author=

Multiple Hypothesis Testing in Conjoint Analysis , volume=. Political Analysis , author=. 2023 , pages=. doi:10.1017/pan.2022.30 , number=

work page doi:10.1017/pan.2022.30 2023
[33]

Journal of Economic Literature , year =

Abadie, Alberto , title =. Journal of Economic Literature , year =

work page
[34]

2017 , OPTkey =

An Evalution of 2016 Election Polls in the. 2017 , OPTkey =

work page 2016
[35]

2017 , OPTkey =

User Guide and Codebook for the. 2017 , OPTkey =

work page 2017
[36]

2017 , OPTkey =

Ansolabehere, Stephen and Schaffner, Brian and Luks, Sam , title =. 2017 , OPTkey =

work page 2017
[37]

R: A Language and Environment for Statistical Computing , Url =

work page
[38]

2004 , OPTkey =

The Health Consequences of Smoking: A Report of the Surgeon General , institution =. 2004 , OPTkey =

work page 2004
[39]

Abadie, Alberto , Journal =

work page
[40]

Abadie, Alberto and Angrist, Joshua and Imbens, Guido , Journal =

work page
[41]

Synthetic Control Methods for Comparative Case Studies: Estimating the Effect of

Abadie, Alberto and Diamond, Alexis and Hainmueller, Jens , Journal =. Synthetic Control Methods for Comparative Case Studies: Estimating the Effect of

work page
[42]

American Economic Review , year =

Abadie, Alberto and Gardeazabal, Javier , title =. American Economic Review , year =

work page
[43]

, Journal =

Abadie, Alberto and Imbens, Guido W. , Journal =

work page
[44]

, Journal =

Abadie, Alberto and Imbens, Guido W. , Journal =. On the Failure of the Bootstrap for Matching Estimators , Volume =

work page
[45]

, Journal =

Abadie, Alberto and Imbens, Guido W. , Journal =. Bias-Corrected Matching Estimators for Average Treatment Effects , Volume =

work page
[46]

, Journal =

Abadie, Alberto and Imbens, Guido W. , Journal =. A Martingale Representation for Matching Estimators , Year =

work page
[47]

, title =

Abramowitz, Alan I. , title =. Journal of Politics , year =

work page
[48]

2018 , OPTkey =

Abraham, Sarah and Sun, Liyang , title =. 2018 , OPTkey =

work page 2018
[49]

, Journal =

Acemoglu, Daron and Johnson, Simon and Robinson, James A. , Journal =

work page
[50]

, Journal =

Achen, Christopher H. , Journal =

work page
[51]

, Publisher =

Achen, Christopher H. , Publisher =

work page
[52]

and Bartels, Larry , Journal =

Achen, Christopher H. and Bartels, Larry , Journal =

work page
[53]

2010 , OPTkey =

Achen, Christopher and Blais, Andre , title =. 2010 , OPTkey =

work page 2010
[54]

and Shively, W

Achen, Christopher H. and Shively, W. Phillips , Publisher =

work page
[55]

and Smith, Dennis J

Adams, Williams C. and Smith, Dennis J. , Journal =

work page
[56]

Adolph, Christopher and King, Gary , Journal =

work page
[57]

and Shotts, Kenneth W

Adolph, Christopher and King, Gary and Herron, Michael C. and Shotts, Kenneth W. , Journal =

work page
[58]

Statistical Methods for the Social Sciences , publisher =

Agresti, Alan and Finlay, Barbara , ALTeditor =. Statistical Methods for the Social Sciences , publisher =. 2008 , OPTkey =

work page 2008
[59]

, title =

Ahlquist, John S. , title =. Political Analysis , year =

work page
[60]

2017 , note =

Ahlquist, John , title =. 2017 , note =

work page 2017
[61]

2013 , OPTkey =

Ahlquist, John and Mayer, Kenneth and Jackman, Simon , title =. 2013 , OPTkey =

work page 2013
[62]

and Mayer, Kenneth R

Ahlquist, John S. and Mayer, Kenneth R. and Jackman, Simon , title =. Election Law Journal , year =

work page
[63]

, Journal =

Aigner, Dennis J. , Journal =

work page
[64]

Mixed membership stochastic blockmodels , Volume =

Airoldi, Edoardo M and Blei, David M and Fienberg, Stephen E and Xing, Eric P , Journal =. Mixed membership stochastic blockmodels , Volume =

work page
[65]

, Publisher =

Aitchison, J. , Publisher =

work page
[66]

, Journal =

Aitchison, J. , Journal =

work page
[67]

and Anderson, D

Aitkin, M. and Anderson, D. and Hinde, J. , Journal =

work page
[68]

and Rubin, Donald B

Aitkin, M. and Rubin, Donald B. , Journal =

work page
[69]

Akaike, Hirotugu , Chapter =

work page
[70]

Akaike, Hirotugu , Journal =

work page
[71]

Akaike, Hirotugu , Journal =. A

work page
[72]

, Journal =

Albert, Jeffrey M. , Journal =

work page
[73]

, Journal =

Albert, James H. , Journal =. Bayesian Estimation of Normal Ogive Item Response Curves Using

work page
[74]

and Chib, Siddhartha , Journal =

Albert, James H. and Chib, Siddhartha , Journal =

work page
[75]

and Chib, Siddhartha , Journal =

Albert, James H. and Chib, Siddhartha , Journal =. Bayes Inference Via

work page
[76]

and Nelson, Suchitra , Journal =

Albert, Jeffrey M. and Nelson, Suchitra , Journal =

work page
[77]

2019 , OPTkey =

Albright, Alex , title =. 2019 , OPTkey =

work page 2019
[78]

, Journal =

Aldrich, John H. , Journal =

work page
[79]

Alesina, Alberto and Rosenthal, Howard , Publisher =

work page
[80]

2014 , publisher=

Our Declaration: A Reading of the Declaration of Independence in Defense of Equality , author=. 2014 , publisher=

work page 2014

Showing first 80 references.