Pith. sign in

REVIEW 2 minor 62 references

Semiparametric Inference for Half-Trek Estimators in Linear Structural Equation Models

T0 review · 0 major / 2 minor · reviewed 2026-06-26 · grok-4.3

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read Half-trek estimators in linear structural equation models have an explicit influence function that yields asymptotic normality and closed-form variances.

desk verdict This paper derives the influence function and closed-form asymptotic variance for half-trek estimators, completing the inferential tools for a known class of causal estimators. read the letter →

arxiv 2606.26931 v1 pith:53MJEVI4 submitted 2026-06-25 stat.ME math.STstat.TH

Leopold Mareis , Nils Sturma , Mathias Drton This is my paper · ORCID
classification stat.MEmath.STstat.TH
keywords half-trekcriterioninfluencefunctionstructuralequationmodelssemiparametricinferencecausalasymptoticnormalitydirectedmixedgraphslatentconfounding
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper derives the semiparametric influence function for the half-trek criterion estimator of structural coefficients in linear models on directed mixed graphs. The function merges the target node's residual with its identification instruments while recursively correcting for uncertainty in earlier stages of the estimation. This produces asymptotic normality whose variance is available in closed form, so that standard errors, confidence regions, and Wald tests become available for the first time. A reader cares because prior graphical identification results stopped at point estimation; the new theory completes the pipeline from covariance matrix to inferential statements about causal effects, including in graphs that contain cycles.

What carries the argument

The semiparametric influence function that combines the structural residual at the target node with identification instruments while recursively correcting for uncertainty from preceding estimation stages.

What would settle it

In repeated simulations drawn from an HTC-identified linear SEM, the empirical coverage of the derived Wald intervals falls materially below the nominal level.

Watch

Extended reading notes

Core claim

We derive the semiparametric influence function of this estimator for all HTC-identified directed mixed graphs, including cyclic ones. The influence function combines the structural residual at the target node with the identification instruments, recursively corrected for uncertainty from earlier estimation stages. The HTC estimator is asymptotically normal with variance computable in closed form, yielding confidence regions, marginal intervals, and Wald tests for individual structural coefficients.

Load-bearing premise

The observed data are generated exactly by a linear structural equation model on a directed mixed graph that satisfies the half-trek criterion and whose covariance matrix admits the rational HTC estimator.

Editorial extensions

If this is right

  • The HTC estimator is asymptotically normal.
  • Its asymptotic variance has a closed-form expression.
  • Valid confidence regions, marginal intervals, and Wald tests can be constructed for individual structural coefficients.
  • The results hold for both acyclic and cyclic graphs.
  • Applied examples produce complete inferential summaries for causal effects.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The same recursive influence-function construction could be adapted to other rational identification methods that rely on covariance entries.
  • The closed-form variance expression opens the door to analytic sample-size calculations for studies that plan to use HTC estimation.
  • Extensions that relax the exact linear-Gaussian assumption could replace the influence function with a robust version while retaining the same graphical skeleton.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit.

Referee Report

0 major / 2 minor

Summary. The paper derives the semiparametric influence function for the rational half-trek criterion (HTC) estimator of structural coefficients in linear structural equation models on directed mixed graphs (including cyclic graphs). The influence function is constructed recursively by combining the structural residual at the target node with identification instruments, with corrections for estimation uncertainty propagated from earlier stages. The resulting estimator is shown to be asymptotically normal with a closed-form asymptotic variance, which is used to construct confidence regions, marginal intervals, and Wald tests. The theory is applied to the Fulton Fish Market dataset to obtain inferential summaries for the causal effect of supply on demand under the maintained assumption that the graph satisfies the HTC.

Significance. If the derivation holds, the work supplies the previously missing asymptotic theory and valid standard errors for HTC estimators, which are closed-form rational functions of the sample covariance. This enables reliable inference for causal effects in the presence of latent confounding whenever the half-trek criterion applies, including on cyclic graphs. The explicit recursive correction for multi-stage estimation and the closed-form variance expression are practical strengths that follow the standard pattern for plug-in estimators whose identification maps are rational functions of the covariance.

minor comments (2)
  1. [§3.3] §3.3: the recursive definition of the influence function (Eq. (12)) would benefit from an explicit low-dimensional worked example (e.g., a three-node cyclic graph) to illustrate how the correction terms are computed in practice.
  2. [§5] §5: the Fulton Fish Market application reports point estimates and intervals but does not include a sensitivity check under mild violations of the linear SEM assumption or a comparison with a non-HTC estimator.

Simulated Author's Rebuttal

0 responses · 0 unresolved

We thank the referee for their careful reading and positive evaluation of the manuscript. The recommendation for minor revision is noted, and we are pleased that the significance of the closed-form asymptotic theory and recursive influence function construction is recognized. Since no specific major comments were raised, we address the overall report below.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity in derivation chain

full rationale

The paper's central contribution is an explicit derivation of the semiparametric influence function for the rational HTC estimator, including recursive corrections for multi-stage estimation, under the maintained graphical identification assumption. This follows the standard pattern for plug-in estimators with rational identification maps and does not reduce any claimed result to a fitted quantity or self-citation by construction. No load-bearing self-citations, self-definitional steps, or ansatz smuggling are visible in the provided abstract or description; the variance expression is presented as newly derived from the influence function rather than presupposed.

Assumptions & free parameters 0 free parameters · 1 assumptions · 0 invented entities

Abstract-only review prevents exhaustive audit. The work rests on standard semiparametric regularity conditions and the graphical identification assumption already stated in the HTC literature.

assumptions (1)
  • domain assumption The observed data are generated by a linear structural equation model whose parameters are identified by the half-trek criterion.
    Invoked to guarantee that the rational estimator exists and the influence function applies.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Semiparametric Inference for Half-Trek Estimators in Linear Structural Equation Models." pith.science (2026). https://pith.science/paper/53MJEVI4

@misc{pith2026260626931,
  author       = {Pith},
  title        = {Pith review of: Semiparametric Inference for Half-Trek Estimators in Linear Structural Equation Models},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/53MJEVI4}},
  note         = {Machine review of arXiv:2606.26931}
}
read the original abstract

Linear structural equation models on directed mixed graphs encode causal relationships among variables subject to latent confounding. The half-trek criterion (HTC) provides a graphical sufficient condition for the structural coefficients to be rationally identifiable from the observable covariance matrix, and yields a corresponding closed-form rational estimator. Despite this, the asymptotic distribution of the HTC estimator, and hence valid standard errors and confidence regions, have not been derived. We derive the semiparametric influence function of this estimator for all HTC-identified directed mixed graphs, including cyclic ones. The influence function combines the structural residual at the target node with the identification instruments, recursively corrected for uncertainty from earlier estimation stages. The HTC estimator is asymptotically normal with variance computable in closed form, yielding confidence regions, marginal intervals, and Wald tests for individual structural coefficients. Applied to the Fulton Fish Market dataset, our theory delivers a complete inferential summary for the causal effect of supply on demand.

Figures

Figures reproduced from arXiv: 2606.26931 by the authors.

Figure 1
Figure 1. Five-node directed mixed graph (cf. Example 6(b) in Foygel et al. (2012)) with HTC ordering 2 ≺ 4 ≺ 5. 3e−04 1e−03 3e−03 1e−02 250 500 1,000 2,000 4,000 8,000 Sample size n Var ( ^ β15 ) 3e−04 1e−03 3e−03 250 500 1,000 2,000 4,000 8,000 Sample size n Var ( ^ β35 ) 1e−04 3e−04 1e−03 3e−03 250 500 1,000 2,000 4,000 8,000 Sample size n |Cov ( ^ β15, ^ β35 )| −0.4 −0.2 0.0 250 500 1,000 2,000 4,000 8,000 Sample size n C… view at source ↗
Figure 2
Figure 2. Calibration of Vˆ 5 for (βˆ 15, βˆ 35) in [PITH_FULL_IMAGE:figures/full_fig_p008_2.png] view at source ↗
Figure 3
Figure 3. Estimation variance of βˆ 12 in [PITH_FULL_IMAGE:figures/full_fig_p009_3.png] view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: summary(htcfit()) output for the cyclic five-node directed mixed graph (left, n = 1000, true β45 = 0); see Appendix C.2 for graph and parameter details. supply wave2 wave3 demand δ HTC Estimator Summary ======================================================= Node deman…
Figure 5
Figure 5. Figure 5: summary(htcfit()) applied to the Fulton Fish Market dataset (n = 97): cyclic directed mixed graph with correlated (supply, demand) errors (left), estimated separately with wave2 and wave3 as instruments for the demand elasticity δ. 11 [PITH_FULL_IMAGE:figures/full_fig…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

62 extracted references

  1. [1]

    Uncertainty in artificial intelligence , pages=

    Identification in missing data models represented by directed acyclic graphs , author=. Uncertainty in artificial intelligence , pages=. 2020 , organization=

    2020

  2. [2]

    , TITLE =

    Tsiatis, Anastasios A. , TITLE =. 2006 , PAGES =

    2006

  3. [3]

    and Robins, James M

    Richardson, Thomas S. and Robins, James M. , TITLE =. Journal of Causal Inference , FJOURNAL =. 2023 , NUMBER =

    2023

  4. [4]

    Eighteenth National Conference on Artificial Intelligence , pages =

    Tian, Jin and Pearl, Judea , title =. Eighteenth National Conference on Artificial Intelligence , pages =. 2002 , publisher =

    2002

  5. [5]

    The Annals of Statistics , FJOURNAL =

    Drton, Mathias and Foygel, Rina and Sullivant, Seth , TITLE =. The Annals of Statistics , FJOURNAL =. 2011 , NUMBER =

    2011

  6. [6]

    The Annals of Statistics , FJOURNAL =

    Foygel, Rina and Draisma, Jan and Drton, Mathias , TITLE =. The Annals of Statistics , FJOURNAL =. 2012 , NUMBER =

    2012

  7. [7]

    Biometrika , FJOURNAL =

    Tan, Zhiqiang , TITLE =. Biometrika , FJOURNAL =. 2010 , NUMBER =

    2010

  8. [8]

    Thirty-Fifth

    Yonghan Jung and Jin Tian and Elias Bareinboim , title =. Thirty-Fifth. 2021 , timestamp =

    2021

Show all 62 references
  1. [9]

    Journal of the American Statistical Association , FJOURNAL =

    Mohan, Karthika and Pearl, Judea , TITLE =. Journal of the American Statistical Association , FJOURNAL =. 2021 , NUMBER =

    2021

  2. [10]

    Scandinavian Journal of Statistics , volume =

    Vansteelandt, Stijn and Didelez, Vanessa , title =. Scandinavian Journal of Statistics , volume =

  3. [11]

    2025 , eprint=

    Automatic debiasing of neural networks via moment-constrained learning , author=. 2025 , eprint=

    2025

  4. [12]

    IEEE Transactions on Signal Processing , volume=

    Active learning and basis selection for kernel-based linear models: A Bayesian perspective , author=. IEEE Transactions on Signal Processing , volume=. 2010 , publisher=

    2010

  5. [13]

    Advances in Neural Information Processing Systems , volume=

    Causal influence detection for improving efficiency in reinforcement learning , author=. Advances in Neural Information Processing Systems , volume=

  6. [14]

    American Economic Review , volume=

    Double/debiased/neyman machine learning of treatment effects , author=. American Economic Review , volume=. 2017 , publisher=

    2017

  7. [15]

    Cognitive Science , volume=

    A Bayesian theory of sequential causal learning and abstract transfer , author=. Cognitive Science , volume=. 2016 , publisher=

    2016

  8. [16]

    Advances in Neural Information Processing Systems , volume=

    Active bayesian causal inference , author=. Advances in Neural Information Processing Systems , volume=

  9. [17]

    Conference on Causal Learning and Reasoning , pages=

    Can Active Sampling Reduce Causal Confusion in Offline Reinforcement Learning? , author=. Conference on Causal Learning and Reasoning , pages=. 2023 , organization=

    2023

  10. [18]

    Journal of Machine Learning Research , year =

    Henrik von Kleist and Alireza Zamanian and Ilya Shpitser and Narges Ahmidi , title =. Journal of Machine Learning Research , year =

  11. [19]

    1977 , publisher=

    Sampling Techniques , author=. 1977 , publisher=

    1977

  12. [20]

    Proceedings of the Conference on Research Data Infrastructure , author=

    MaRDI's Zenodo Community for Graphical Modeling and Causal Inference , volume=. Proceedings of the Conference on Research Data Infrastructure , author=

  13. [21]

    Causal protein-signaling networks derived from multiparameter single-cell data

    Sachs, Karen and Perez, Omar and Pe'er, Dana and Lauffenburger, Douglas A and Nolan, Garry P. Causal protein-signaling networks derived from multiparameter single-cell data. Science

  14. [22]

    MIMIC-IV , a freely accessible electronic health record dataset

    Johnson, Alistair E W and Bulgarelli, Lucas and Shen, Lu and Gayles, Alvin and Shammout, Ayad and Horng, Steven and Pollard, Tom J and Hao, Sicheng and Moody, Benjamin and Gow, Brian and Lehman, Li-Wei H and Celi, Leo A and Mark, Roger G. MIMIC-IV , a freely accessible electro...

  15. [23]

    2024 , volume =

    G\"obler, Konstantin and Windisch, Tobias and Drton, Mathias and Pychynski, Tim and Roth, Martin and Sonntag, Steffen , booktitle =. 2024 , volume =

    2024

  16. [24]

    Statistical Science , FJOURNAL =

    van der Vaart, Aad , TITLE =. Statistical Science , FJOURNAL =. 2014 , NUMBER =

    2014

  17. [25]

    , TITLE =

    van der Vaart, Aad W. , TITLE =. 1998 , PAGES =

    1998

  18. [26]

    and van der Laan, Mark J

    Gill, Richard D. and van der Laan, Mark J. and Robins, James M. and Fleming, T. R. Coarsening at Random: Characterizations, Conjectures, Counter-Examples. Proceedings of the First Seattle Symposium in Biostatistics. 1997

    1997

  19. [27]

    2008 , PAGES =

    Ito, Kazufumi and Kunisch, Karl , TITLE =. 2008 , PAGES =

    2008

  20. [28]

    The American Statistician , FJOURNAL =

    Hines, Oliver and Dukes, Oliver and Diaz-Ordaz, Karla and Vansteelandt, Stijn , TITLE =. The American Statistician , FJOURNAL =. 2022 , NUMBER =

    2022

  21. [29]

    Maathuis and Vanessa Didelez , title =

    Janine Witte and Leonard Henckel and Marloes H. Maathuis and Vanessa Didelez , title =. Journal of Machine Learning Research , year =

  22. [30]

    and Rotnitzky, Andrea and Zhao, Lue Ping , TITLE =

    Robins, James M. and Rotnitzky, Andrea and Zhao, Lue Ping , TITLE =. Journal of the American Statistical Association , FJOURNAL =. 1994 , NUMBER =

    1994

  23. [31]

    and Rotnitzky, Andrea , TITLE =

    Robins, James M. and Rotnitzky, Andrea , TITLE =. Journal of the American Statistical Association , FJOURNAL =. 1995 , NUMBER =

    1995

  24. [32]

    Journal of Machine Learning Research , year =

    Rohit Bhattacharya and Razieh Nabi and Ilya Shpitser , title =. Journal of Machine Learning Research , year =

  25. [33]

    Optimizing Experimental Design for Causal Effect Estimation with Partial Measurements

    Mareis, Leopold. Optimizing Experimental Design for Causal Effect Estimation with Partial Measurements. Causal Inference. 2025

    2025

  26. [34]

    Newey , journal =

    Whitney K. Newey , journal =. Semiparametric Efficiency Bounds , volume =

  27. [35]

    2019 , PAGES =

    Handbook of. 2019 , PAGES =

    2019

  28. [36]

    , title =

    Henckel, Leonard and Buttenschoen, Martin and Maathuis, Marloes H. , title =. Biometrika , volume =

  29. [37]

    , title =

    Henckel, Leonard and Perković, Emilija and Maathuis, Marloes H. , title =. Journal of the Royal Statistical Society Series B: Statistical Methodology , volume =

  30. [38]

    1987 , author =

    Asymptotic efficiency in estimation with conditional moment restrictions , journal =. 1987 , author =

    1987

  31. [39]

    Newey , journal =

    Whitney K. Newey , journal =. Efficient

  32. [40]

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

  33. [41]

    1928 , publisher=

    The Tariff on Animal and Vegetable Oils , author=. 1928 , publisher=

    1928

  34. [42]

    and Turkington, Darrell A

    Bowden, Roger J. and Turkington, Darrell A. , year=. Instrumental Variables , publisher=

  35. [43]

    Proceedings of the Twenty-Second Conference on Uncertainty in Artificial Intelligence , pages =

    Brito, Carlos and Pearl, Judea , title =. Proceedings of the Twenty-Second Conference on Uncertainty in Artificial Intelligence , pages =. 2006 , publisher =

    2006

  36. [44]

    Proceedings of the 21st International Joint Conference on Artificial Intelligence , pages =

    Tian, Jin , title =. Proceedings of the 21st International Joint Conference on Artificial Intelligence , pages =. 2009 , publisher =

    2009

  37. [45]

    Identification and

    Bryant Chen and Daniel Kumor and Elias Bareinboim , booktitle =. Identification and. 2017 , volume =

    2017

  38. [46]

    Efficient

    Kumor, Daniel and Chen, Bryant and Bareinboim, Elias , booktitle =. Efficient

  39. [47]

    Scandinavian Journal of Statistics , FJOURNAL =

    Drton, Mathias and Weihs, Luca , TITLE =. Scandinavian Journal of Statistics , FJOURNAL =. 2016 , NUMBER =

    2016

  40. [48]

    Journal of Causal Inference , FJOURNAL =

    Weihs, Luca and Robinson, Bill and Dufresne, Emilie and Kenkel, Jennifer and Kubjas, Kaie and McGee, II, Reginald and Nguyen, Nhan and Robeva, Elina and Drton, Mathias , TITLE =. Journal of Causal Inference , FJOURNAL =. 2018 , NUMBER =

    2018

  41. [49]

    Proceedings of the Twenty-Fifth International Joint Conference on Artificial Intelligence , pages =

    Chen, Bryant and Pearl, Judea and Bareinboim, Elias , title =. Proceedings of the Twenty-Fifth International Joint Conference on Artificial Intelligence , pages =. 2016 , publisher =

    2016

  42. [50]

    SIAM Journal on Matrix Analysis and Applications , FJOURNAL =

    Dettling, Philipp and Homs, Roser and Am\'endola, Carlos and Drton, Mathias and Hansen, Niels Richard , TITLE =. SIAM Journal on Matrix Analysis and Applications , FJOURNAL =. 2023 , NUMBER =

    2023

  43. [51]

    The Annals of Statistics , FJOURNAL =

    Barber, Rina Foygel and Drton, Mathias and Sturma, Nils and Weihs, Luca , TITLE =. The Annals of Statistics , FJOURNAL =. 2022 , NUMBER =

    2022

  44. [52]

    Graphical continuous

    Varando, Gherardo and Richard Hansen, Niels , booktitle =. Graphical continuous. 2020 , volume =

    2020

  45. [53]

    2026 , Eprint =

    Benjamin Hollering and Pratik Misra and Nils Sturma , Title =. 2026 , Eprint =

    2026

  46. [54]

    2026 , Eprint =

    Leopold Mareis and Mathias Drton , Title =. 2026 , Eprint =

    2026

  47. [55]

    2018 , PAGES =

    Sullivant, Seth , TITLE =. 2018 , PAGES =

    2018

  48. [56]

    , TITLE =

    Bollen, Kenneth A. , TITLE =. 1989 , PAGES =

    1989

  49. [57]

    2000 , PAGES =

    Spirtes, Peter and Glymour, Clark and Scheines, Richard , TITLE =. 2000 , PAGES =

    2000

  50. [58]

    and Graddy, Kathryn and Imbens, Guido W

    Angrist, Joshua D. and Graddy, Kathryn and Imbens, Guido W. , title =. The Review of Economic Studies , volume =

  51. [59]

    Testing for

    Kathryn Graddy , journal =. Testing for

  52. [60]

    2025 , note =

    SEMID: Identifiability of Linear Structural Equation Models , author =. 2025 , note =

    2025

  53. [61]

    Identifying causal effects with computer algebra , year =

    Garc\'. Identifying causal effects with computer algebra , year =. Proceedings of the Twenty-Sixth Conference on Uncertainty in Artificial Intelligence , pages =

  54. [62]

    Proceedings of the AAAI Conference on Artificial Intelligence , author=

    Estimating Identifiable Causal Effects through Double Machine Learning , volume=. Proceedings of the AAAI Conference on Artificial Intelligence , author=. 2021 , month=

    2021

Pith tools

Reviewed June 26, 2026 · model on record in the stance chip above.