pith. sign in

arxiv: 2605.16593 · v1 · pith:6GVDWK3Wnew · submitted 2026-05-15 · 📊 stat.AP · econ.EM· stat.ML

Policy Learning with Observational Data: The Case of Hepatitis C Treatment for HIV/HCV Co-Infected Patients

Rapha\"el Langevin This is my paper

Pith reviewed 2026-05-19 20:53 UTC · model grok-4.3

classification 📊 stat.AP econ.EMstat.ML
keywords policy learningobservational dataconditional average treatment effectshepatitis CHIV co-infectiondecision treestreatment allocationcost savings
0
0 comments X

The pith

Reallocating hepatitis C treatments among HIV co-infected patients could cut costs by CAN$3.6-4.9 million while increasing health benefits.

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

The paper develops a general method to derive policy rules from observational data for choosing among multiple treatment options when patients differ in their responses. It estimates conditional average treatment effects consistently by applying a weighted K-means algorithm to partition the sample into homogeneous subgroups where an outcome model holds, then converts those estimates into practical rules using a standard decision tree that accommodates both full and partial adherence. Applied to modern therapies for hepatitis C in people also living with HIV, where no single guideline exists, the approach identifies one subgroup with roughly an 80 percent chance of clearing the virus without any drug. It further shows that shifting which treated patients receive which therapy would lower total costs by CAN$3.6-4.9 million while raising overall health gains relative to current patterns.

Core claim

Using observational data, the paper derives policy rules that reallocate modern HCV treatments among treated HIV/HCV co-infected patients. This reallocation reduces total treatment costs by CAN$3.6-4.9 million while increasing aggregate health benefits relative to the status quo. The method also identifies a subgroup of patients with approximately an 80 percent probability of spontaneous HCV clearance without treatment.

What carries the argument

Weighted K-means algorithm that partitions patients into homogeneous subgroups for consistent estimation of conditional average treatment effects, followed by a decision tree that translates those effects into feasible policy rules allowing for perfect or imperfect adherence.

Load-bearing premise

The outcome model is correctly specified within each homogeneous subgroup identified by the weighted K-means algorithm.

What would settle it

A randomized trial that assigns patients according to the derived policy rules and finds no reduction in total costs or no increase in aggregate health benefits compared with current practice would falsify the central claim.

Figures

Figures reproduced from arXiv: 2605.16593 by Rapha\"el Langevin.

Figure 1
Figure 1. Figure 1: Schematic representation of the different objects used to learn policy rules. logic of Chernozhukov et al. (2025). The estimated CATEs are then used to derive both infea￾sible and feasible policy rules. Finally, adherence to treatment is modeled via a two-part model where predicted adherence is then incorporated into both policy rules to account for variations in adherence across individuals and treatment … view at source ↗
Figure 2
Figure 2. Figure 2: Geographical distribution of the participants enrolled in the Canadian Co-infection Cohort, 2003-2023 patients were approached to participate in order to avoid selection bias (Klein et al., 2010). A total of 19 clinical sites are participating in the CCC as of the end of 2023 [PITH_FULL_IMAGE:figures/full_fig_p026_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Socioeconomic characteristics at enrollment, Canadian Co-Infection Cohort, 2003-2023 injection drug use, low-income Indigenous people are at increased risk of being co-infected. In Canada, evidence suggests that Indigenous peoples account for 70% to 80% of new HepC infections among individuals who inject drugs (Saeed et al., 2024). 4.2 HIV, HCV, and Direct-Acting Antiviral Agents Around 6 million people ar… view at source ↗
Figure 4
Figure 4. Figure 4: Estimated median coefficients and their respective confidence intervals obtained from the LPMs for each group and for the full sample when G = 5 and λmin = 0.7. spontaneous clearance for individuals within this group in the target population. This might not be the case in practice, as there are significant costs to society associated with not treating a patient who may spread the infection while waiting. T… view at source ↗
Figure 5
Figure 5. Figure 5: Estimated median coefficients and their respective confidence intervals obtained from the LPMs for each group and for the full sample when G = 3 and λmin = 0.04. For comparison purposes, [PITH_FULL_IMAGE:figures/full_fig_p038_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Results of the cost-effectiveness analysis with perfect adherence to treatment and estimated group memberships zˆ(µˆ, Σ). ˆ the aggregate level. To give a better sense of the magnitude of the health benefits compared to the cost of treatment, it is possible to compare the total cost of a given treatment option per expected gains in quality-adjusted life years (QALYs). For instance, if we assume that Mavyre… view at source ↗
Figure 7
Figure 7. Figure 7: Results of the cost-effectiveness analysis with perfect adherence to treatment and tree-based predicted group memberships hˆ(V ∗ ). perfect adherence, and that lower adherence always leads to lower health benefits, the aggregate ICER shown in Panel (c) of [PITH_FULL_IMAGE:figures/full_fig_p044_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Results of the cost-effectiveness analysis with predicted adherence to treatment and estimated group memberships zˆ(µˆ, Σ). ˆ Finally, [PITH_FULL_IMAGE:figures/full_fig_p045_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Results of the cost-effectiveness analysis with predicted adherence to treatment and tree￾based predicted group memberships hˆ(V ∗ ). when compared to the status quo. When WTP is below CAN$900/pp, both total costs and health benefits are negative compared to the status quo allocation (with predicted adherence), whereas total costs and health benefits become both positive if WTP > CAN$900/pp. Finally, Panel… view at source ↗
Figure 10
Figure 10. Figure 10: Selected decision tree for the feasible policy rule. 73 [PITH_FULL_IMAGE:figures/full_fig_p074_10.png] view at source ↗
read the original abstract

Decision-makers frequently must choose a single action from a finite set of alternatives -- for example, physicians selecting a treatment, investors choosing a portfolio risk level, or judges determining sentences. To improve outcomes, policymakers often issue policy rules or guidelines to inform such choices. In this paper, I show how to generally derive policy rules from observational data in a multi-action framework under relatively weak assumptions about the underlying structure of the heterogeneous sampled population. Conditional average treatment effects (CATEs) are consistently estimated via a weighted K-means algorithm, assuming the outcome model is correctly specified within each homogeneous subgroup. Feasible policy rules are then implemented via a standard decision tree, allowing for both perfect and imperfect adherence to treatment. The methodology is applied to treatment options for Hepatitis C (HCV) among patients co-infected with human immunodeficiency virus (HIV), a setting in which no uniform guideline exists for modern pharmaceutical therapies. The results identify a subgroup of patients with approximately an 80% probability of spontaneous HCV clearance without treatment. Estimation results also show that reallocating treatments among treated individuals could have reduced total treatment costs by CAN$3.6-4.9 million while still increasing aggregate health benefits relative to the status quo. These findings demonstrate that the proposed approach can generate improved, data-driven treatment guidelines for the management of HIV/HCV co-infected patients.

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.

Referee Report

2 major / 1 minor

Summary. The paper develops a general method to derive feasible policy rules from observational data in multi-action settings by estimating conditional average treatment effects (CATEs) via a weighted K-means algorithm that identifies homogeneous subgroups, under the assumption that the outcome model is correctly specified within each cluster. Feasible policies are then implemented with a standard decision tree that accommodates both perfect and imperfect adherence. The methodology is applied to Hepatitis C treatment choices among HIV/HCV co-infected patients, where no uniform guideline exists; the results identify a subgroup with approximately 80% probability of spontaneous HCV clearance without treatment and claim that reallocating treatments among treated individuals could reduce total treatment costs by CAN$3.6-4.9 million while increasing aggregate health benefits relative to the status quo.

Significance. If the results hold, the paper offers a practical framework for policy learning from observational data under relatively weak structural assumptions on population heterogeneity, which could support data-driven treatment guidelines in clinical settings lacking consensus protocols. The empirical application to HIV/HCV co-infection demonstrates the potential for simultaneous cost reduction and health improvement through reallocation, providing a concrete illustration of how CATE-based policies might inform resource allocation in healthcare.

major comments (2)
  1. [Abstract] Abstract (paragraph on CATE estimation): The claim that CATEs are 'consistently estimated' rests on the assumption that the outcome model is correctly specified within each homogeneous subgroup produced by the weighted K-means algorithm, yet the manuscript provides no diagnostic evidence, model checks, or sensitivity analyses for this assumption (e.g., to the choice of K or weighting scheme). This assumption is load-bearing for the central quantitative claim of CAN$3.6-4.9 million in cost savings, because any systematic bias in the subgroup-specific predictions would directly invalidate the reported reallocation benefits.
  2. [Abstract] Abstract (results on reallocation): The reported cost savings and health-benefit gains are obtained by applying the estimated CATEs to re-assign treatments within the observed sample; the paper does not show that these policy recommendations remain stable under alternative functional forms for the cluster-specific outcome model, which creates a potential circularity between the clustering step and the final policy evaluation.
minor comments (1)
  1. [Abstract] The abstract refers to 'relatively weak assumptions about the underlying structure of the heterogeneous sampled population' without enumerating them explicitly; a brief list or reference to the relevant section would improve clarity for readers.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their constructive comments, which have helped us improve the robustness of our analysis. We address each major comment in turn below, and have made revisions to the manuscript accordingly.

read point-by-point responses

  1. Referee: [Abstract] Abstract (paragraph on CATE estimation): The claim that CATEs are 'consistently estimated' rests on the assumption that the outcome model is correctly specified within each homogeneous subgroup produced by the weighted K-means algorithm, yet the manuscript provides no diagnostic evidence, model checks, or sensitivity analyses for this assumption (e.g., to the choice of K or weighting scheme). This assumption is load-bearing for the central quantitative claim of CAN$3.6-4.9 million in cost savings, because any systematic bias in the subgroup-specific predictions would directly invalidate the reported reallocation benefits.

    Authors: We agree that the consistency claim depends on correct specification within clusters, and that additional checks are warranted. The weighted K-means is designed to identify subgroups where a common outcome model applies, but we recognize the need for empirical validation. In the revised manuscript, we include sensitivity analyses to the choice of K (testing K=3 to K=6) and alternative weighting schemes. We also report within-cluster goodness-of-fit measures and residual plots to support the model specification. These new results confirm that the main findings, including the identification of the high-clearance subgroup, are robust to these variations. revision: yes

  2. Referee: [Abstract] Abstract (results on reallocation): The reported cost savings and health-benefit gains are obtained by applying the estimated CATEs to re-assign treatments within the observed sample; the paper does not show that these policy recommendations remain stable under alternative functional forms for the cluster-specific outcome model, which creates a potential circularity between the clustering step and the final policy evaluation.

    Authors: The referee raises a valid point regarding potential circularity in the in-sample policy evaluation. To mitigate this concern, we have added analyses using alternative functional forms for the outcome models within clusters, such as linear probability models versus logistic regression, and re-computed the reallocation benefits. The revised results show that the direction of the cost savings (CAN$3.6-4.9 million range) and the health benefits remain consistent, although the precise figures vary slightly with the specification. We have updated the abstract and main text to note that these are in-sample estimates and discuss the implications for policy stability. We acknowledge that fully out-of-sample validation would require additional data not available in the current study. revision: partial

Circularity Check

0 steps flagged

No circularity: derivation relies on explicit modeling assumptions and standard estimation steps

full rationale

The paper estimates CATEs via weighted K-means under the stated assumption that the outcome model is correctly specified within each homogeneous subgroup, then derives feasible policy rules via decision tree and applies them to compute reallocation effects on costs and benefits. This chain is an empirical procedure whose quantitative outputs depend on the validity of the modeling assumptions and data, but does not reduce any result to its inputs by definition, by renaming a fit as a prediction, or by self-citation load-bearing. No equations or steps in the provided text exhibit the required reduction (e.g., Eq. X = Eq. Y by construction). The approach is therefore self-contained against external benchmarks once the assumptions are granted.

Axiom & Free-Parameter Ledger

2 free parameters · 1 axioms · 0 invented entities

The central claim rests on the assumption that the outcome model is correctly specified inside each cluster and on the implicit choice of the number of clusters and the weighting scheme used in K-means. No new entities are postulated.

free parameters (2)
  • number of clusters K
    Chosen to define homogeneous subgroups for the outcome model; value not reported in abstract.
  • cluster-specific outcome model parameters
    Fitted inside each K-means group; correctness of these fits is required for consistent CATE estimation.
axioms (1)
  • domain assumption Outcome model is correctly specified within each homogeneous subgroup
    Stated in the abstract as the condition for consistent CATE estimation.

pith-pipeline@v0.9.0 · 5775 in / 1383 out tokens · 33851 ms · 2026-05-19T20:53:17.716712+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Lean theorems connected to this paper

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

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

299 extracted references · 299 canonical work pages

  1. [1]

    Journal of Pharmaceutical Policy and Practice , author =

    Estimating proportion of days covered (. Journal of Pharmaceutical Policy and Practice , author =. 2021 , pages =. doi:10.1186/s40545-021-00385-w , abstract =

    work page doi:10.1186/s40545-021-00385-w 2021

  2. [2]

    Health Economics , author =

    Therapeutic non‐adherence: a rational behavior revealing patient preferences? , volume =. Health Economics , author =. 2007 , pages =. doi:10.1002/hec.1214 , abstract =

    work page doi:10.1002/hec.1214 2007

  3. [3]

    Langevin, Raphaël , month = feb, year =. Bias-. doi:10.48550/arXiv.2601.20197 , abstract =

    work page doi:10.48550/arxiv.2601.20197

  4. [4]

    Journal of Managed Care & Specialty Pharmacy , author =

    Changes in. Journal of Managed Care & Specialty Pharmacy , author =. 2020 , pages =. doi:10.18553/jmcp.2020.26.7.879 , abstract =

    work page doi:10.18553/jmcp.2020.26.7.879 2020

  5. [5]

    Generalized

    Tibshirani, Julie and Athey, Susan and Sverdrup, Erik and Wager, Stefan , month = nov, year =. Generalized

    work page

  6. [6]

    Value in Health , author =

    Making. Value in Health , author =. 2017 , note =. doi:10.1016/j.jval.2017.08.3012 , language =

    work page doi:10.1016/j.jval.2017.08.3012 2017

  7. [7]

    Bay, Yong Yi and Yearick, Kathleen A , year =. Machine

    work page

  8. [8]

    Addiction (Abingdon, England) , author =

    Generalizability of. Addiction (Abingdon, England) , author =. 2017 , pmid =. doi:10.1111/add.13789 , abstract =

    work page doi:10.1111/add.13789 2017

  9. [9]

    Medical Care , author =

    Creating and. Medical Care , author =. 2007 , note =. doi:10.1097/mlr.0b013e3180616c3f , abstract =

    work page doi:10.1097/mlr.0b013e3180616c3f 2007

  10. [10]

    Econometrica , author =

    Root-. Econometrica , author =. 1988 , note =. doi:10.2307/1912705 , abstract =

    work page doi:10.2307/1912705 1988

  11. [11]

    and Sun, Liyang , month = feb, year =

    Chernozhukov, Victor and Lee, Sokbae and Rosen, Adam M. and Sun, Liyang , month = feb, year =. Policy. doi:10.48550/arXiv.2502.10653 , abstract =

    work page doi:10.48550/arxiv.2502.10653

  12. [12]

    Construction of

    Zhao, Wei and Jiang, Xuehan and Wang, Ke and Sun, Xingzhi and Hu, Gang and Xie, Guotong , year =. Construction of. Studies in. doi:10.3233/shti200015 , note =

    work page doi:10.3233/shti200015

  13. [13]

    Journal of Medical Systems , author =

    Decision. Journal of Medical Systems , author =. 2002 , file =

    work page 2002

  14. [14]

    Econometrica , author =

    Fisher-. Econometrica , author =. 2025 , pages =

    work page 2025

  15. [15]

    Journal of the American Statistical Association , author =

    Finding the. Journal of the American Statistical Association , author =. 2003 , note =. doi:10.1198/016214503000000666 , abstract =

    work page doi:10.1198/016214503000000666 2003

  16. [16]

    Journal of the Royal Statistical Society Series B: Statistical Methodology , author =

    Estimating the. Journal of the Royal Statistical Society Series B: Statistical Methodology , author =. 2001 , pages =. doi:10.1111/1467-9868.00293 , abstract =

    work page doi:10.1111/1467-9868.00293 2001

  17. [17]

    European Radiology , author =

    Disadvantages of using the area under the receiver operating characteristic curve to assess imaging tests:. European Radiology , author =. 2015 , pmid =. doi:10.1007/s00330-014-3487-0 , abstract =

    work page doi:10.1007/s00330-014-3487-0 2015

  18. [18]

    International Journal of Epidemiology , author =

    Reflection on modern methods:. International Journal of Epidemiology , author =. 2020 , pages =. doi:10.1093/ije/dyz274 , abstract =

    work page doi:10.1093/ije/dyz274 2020

  19. [19]

    and Olshen, Richard A

    Breiman, Leo and Friedman, Jerome H. and Olshen, Richard A. and Stone, Charles J. , month = oct, year =. Classification

    work page

  20. [20]

    Canada’s

    Government of Canada, Statistics Canada , month = jun, year =. Canada’s

    work page

  21. [21]

    doi:10.25318/1110019201-ENG , urldate =

    Table 11-10-0192-01,. doi:10.25318/1110019201-ENG , urldate =

    work page doi:10.25318/1110019201-eng

  22. [22]

    Management Science , author =

    Policy. Management Science , author =. 2024 , pages =. doi:10.1287/mnsc.2023.4921 , abstract =

    work page doi:10.1287/mnsc.2023.4921 2024

  23. [23]

    Viral Hepatitis , author =

    Viral. Viral Hepatitis , author =. 2025 , file =

    work page 2025

  24. [24]

    National Institute of Health , month = mar, year =

    work page

  25. [25]

    Including non-randomized studies on intervention effects , booktitle =

    Reeves, Barnaby C and Deeks, Jonathan J and Higgins, Julian PT and Shea, Beverley and Tugwell, Peter and Wells, George A and on behalf of the Cochrane Non-Randomized Studies of Interventions Methods Group , publisher =. Including non-randomized studies on intervention effects , booktitle =. doi:https://doi.org/10.1002/9781119536604.ch24 , url =. https://o...

    work page doi:10.1002/9781119536604.ch24

  26. [26]

    Draft Guidance for Industry and Food and Drug Administration Staff , author =

    Use of. Draft Guidance for Industry and Food and Drug Administration Staff , author =. 2023 , pages =

    work page 2023

  27. [27]

    Injury , author =

    Has anything changed in. Injury , author =. 2023 , pages =. doi:10.1016/j.injury.2022.04.012 , language =

    work page doi:10.1016/j.injury.2022.04.012 2023

  28. [28]

    International Journal for Quality in Health Care , author =

    Guide to clinical practice guidelines: the current state of play , volume =. International Journal for Quality in Health Care , author =. 2016 , pages =. doi:10.1093/intqhc/mzv115 , abstract =

    work page doi:10.1093/intqhc/mzv115 2016

  29. [29]

    BMC Health Services Research , author =

    Approaches to clinical guideline development in healthcare: a scoping review and document analysis , volume =. BMC Health Services Research , author =. 2023 , pages =. doi:10.1186/s12913-022-08975-3 , abstract =

    work page doi:10.1186/s12913-022-08975-3 2023

  30. [30]

    1990 , doi =

    Clinical. 1990 , doi =

    work page 1990

  31. [31]

    2021 , pages =

    American Journal of Gastroenterology , author =. 2021 , pages =. doi:10.14309/ajg.0000000000001036 , abstract =

    work page doi:10.14309/ajg.0000000000001036 2021

  32. [32]

    Prevention and. U.S. Centers for Disease Control and Prevention , author =. 2024 , file =

    work page 2024

  33. [33]

    Working Paper , author =

    Understanding and. Working Paper , author =. 2022 , pages =

    work page 2022

  34. [34]

    Neural Networks , author =

    Deterministic annealing. Neural Networks , author =. 1998 , pages =. doi:10.1016/S0893-6080(97)00133-0 , abstract =

    work page doi:10.1016/s0893-6080(97)00133-0 1998

  35. [35]

    Deterministic

    Ueda, Naonori and Nakano, Ryohei , year =. Deterministic. Advances in

    work page

  36. [36]

    Kwon, Jeongyeol and Caramanis, Constantine , month = jun, year =. The

    work page

  37. [37]

    Kwon, Jeongyeol and Qian, Wei and Caramanis, Constantine and Chen, Yudong and Davis, Damek , month = jun, year =. Global. Proceedings of the

    work page

  38. [38]

    Kwon, Jeongyeol and Ho, Nhat and Caramanis, Constantine , month = mar, year =. On the. Proceedings of

    work page

  39. [39]

    Kwon, Jeongyeol and Caramanis, Constantine , month = nov, year =

    work page

  40. [40]

    and Jordan, Michael , month = sep, year =

    Jin, Chi and Zhang, Yuchen and Balakrishnan, Sivaraman and Wainwright, Martin J. and Jordan, Michael , month = sep, year =. Local

    work page

  41. [41]

    Structures of

    Qian, Wei and Zhang, Yuqian and Chen, Yudong , month = feb, year =. Structures of

    work page

  42. [42]

    Likelihood

    Chen, Yudong and Xi, Xumei , month = sep, year =. Likelihood

    work page

  43. [43]

    IEEE Transactions on Information Theory , author =

    Structures of. IEEE Transactions on Information Theory , author =. 2022 , note =. doi:10.1109/TIT.2021.3122465 , abstract =

    work page doi:10.1109/tit.2021.3122465 2022

  44. [44]

    Journal of Econometrics , author =

    Grouped effects estimators in fixed effects models , volume =. Journal of Econometrics , author =. 2016 , pages =. doi:10.1016/j.jeconom.2012.08.022 , abstract =

    work page doi:10.1016/j.jeconom.2012.08.022 2016

  45. [45]

    Econometrica , author =

    Grouped. Econometrica , author =. 2015 , pages =. doi:10.3982/ECTA11319 , language =

    work page doi:10.3982/ecta11319 2015

  46. [46]

    Econometrica , author =

    Grouped. Econometrica , author =. 2015 , note =

    work page 2015

  47. [48]

    Discretizing

    Manresa, Stéphane Bonhomme Thibaut Lamadon Elena , month = feb, year =. Discretizing

    work page

  48. [49]

    Journal of Econometrics , author =

    Heterogeneous structural breaks in panel data models , volume =. Journal of Econometrics , author =. 2021 , pages =. doi:10.1016/j.jeconom.2020.04.009 , abstract =

    work page doi:10.1016/j.jeconom.2020.04.009 2021

  49. [50]

    Econometrica , author =

    Identifying. Econometrica , author =. 2016 , note =. doi:10.3982/ECTA12560 , abstract =

    work page doi:10.3982/ecta12560 2016

  50. [51]

    Journal of Econometrics , author =

    Shrinkage estimation of common breaks in panel data models via adaptive group fused. Journal of Econometrics , author =. 2016 , pages =. doi:10.1016/j.jeconom.2015.09.004 , abstract =

    work page doi:10.1016/j.jeconom.2015.09.004 2016

  51. [52]

    Journal of Applied Econometrics , author =

    To pool or not to pool:. Journal of Applied Econometrics , author =. 2019 , note =. doi:10.1002/jae.2696 , abstract =

    work page doi:10.1002/jae.2696 2019

  52. [53]

    Quantitative Economics , author =

    Determining the number of groups in latent panel structures with an application to income and democracy , volume =. Quantitative Economics , author =. 2017 , note =. doi:10.3982/QE517 , abstract =

    work page doi:10.3982/qe517 2017

  53. [54]

    Journal of Applied Econometrics , author =

    Homogeneity pursuit in panel data models:. Journal of Applied Econometrics , author =. 2018 , note =. doi:10.1002/jae.2632 , abstract =

    work page doi:10.1002/jae.2632 2018

  54. [55]

    Wahed and Peter F

    Panel. Journal of the American Statistical Association , author =. 2016 , pages =. doi:10.1080/01621459.2015.1119696 , abstract =

    work page doi:10.1080/01621459.2015.1119696 2016

  55. [56]

    Journal of Business & Economic Statistics , author =

    Estimation of. Journal of Business & Economic Statistics , author =. 2022 , pages =. doi:10.1080/07350015.2022.2067546 , abstract =

    work page doi:10.1080/07350015.2022.2067546 2022

  56. [57]

    The Econometrics Journal , author =

    Estimating latent group structure in time-varying coefficient panel data models , volume =. The Econometrics Journal , author =. 2019 , pages =. doi:10.1093/ectj/utz008 , abstract =

    work page doi:10.1093/ectj/utz008 2019

  57. [58]

    Journal of Business & Economic Statistics , author =

    Sieve. Journal of Business & Economic Statistics , author =. 2019 , pages =. doi:10.1080/07350015.2017.1340299 , language =

    work page doi:10.1080/07350015.2017.1340299 2019

  58. [59]

    Confidence set for group membership , url =

    Dzemski, Andreas and Okui, Ryo , month = aug, year =. Confidence set for group membership , url =

    work page

  59. [60]

    Journal of Econometrics , author =

    Estimation of panel group structure models with structural breaks in group memberships and coefficients , issn =. Journal of Econometrics , author =. 2022 , pages =. doi:10.1016/j.jeconom.2022.01.001 , abstract =

    work page doi:10.1016/j.jeconom.2022.01.001 2022

  60. [61]

    Journal of the Royal Statistical Society

    Maximum. Journal of the Royal Statistical Society. Series B (Methodological) , author =. 1977 , note =

    work page 1977

  61. [62]

    The Annals of Statistics , author =

    On the. The Annals of Statistics , author =. 1983 , note =

    work page 1983

  62. [63]

    Grouping and clustering methods in econometrics , url =

    Okui, Ryo , year =. Grouping and clustering methods in econometrics , url =

    work page

  63. [64]

    K-means clustering and

    Russo, Nicolò , pages =. K-means clustering and

    work page

  64. [65]

    Econometrica , author =

    A. Econometrica , author =. 2019 , note =. doi:10.3982/ECTA15722 , abstract =

    work page doi:10.3982/ecta15722 2019

  65. [66]

    Scandinavian Journal of Statistics , author =

    Strong. Scandinavian Journal of Statistics , author =. 2009 , note =

    work page 2009

  66. [67]

    SIAM Review , author =

    Mixture. SIAM Review , author =. 1984 , note =

    work page 1984

  67. [68]

    The Annals of Mathematical Statistics , author =

    Note on the. The Annals of Mathematical Statistics , author =. 1949 , note =

    work page 1949

  68. [69]

    The Annals of Statistics , author =

    Note on the. The Annals of Statistics , author =. 1981 , note =

    work page 1981

  69. [70]

    Asymptotic properties of the

    Nityasuddhi, Dechavudh and Bohning, Dankmar , pages =. Asymptotic properties of the

    work page

  70. [71]

    Applied Mathematics-A Journal of Chinese Universities , author =

    Asymptotic properties and expectation-maximization algorithm for maximum likelihood estimates of the parameters from. Applied Mathematics-A Journal of Chinese Universities , author =. 2016 , pages =. doi:10.1007/s11766-016-3391-2 , abstract =

    work page doi:10.1007/s11766-016-3391-2 2016

  71. [72]

    Pattern Recognition , author =

    Gaussian parsimonious clustering models , volume =. Pattern Recognition , author =. 1995 , pages =. doi:10.1016/0031-3203(94)00125-6 , abstract =

    work page doi:10.1016/0031-3203(94)00125-6 1995

  72. [73]

    Journal of Classification , author =

    Clustering criteria for discrete data and latent class models , volume =. Journal of Classification , author =. 1991 , pages =. doi:10.1007/BF02616237 , abstract =

    work page doi:10.1007/bf02616237 1991

  73. [74]

    Journal of Statistical Computation and Simulation , author =

    Comparison of the mixture and the classification maximum likelihood in cluster analysis , volume =. Journal of Statistical Computation and Simulation , author =. 1993 , pages =. doi:10.1080/00949659308811525 , language =

    work page doi:10.1080/00949659308811525 1993

  74. [75]

    The Econometrics Journal , author =

    Using mixtures in econometric models: a brief review and some new results , volume =. The Econometrics Journal , author =. 2016 , note =. doi:10.1111/ectj.12068 , abstract =

    work page doi:10.1111/ectj.12068 2016

  75. [76]

    Computational Statistics & Data Analysis , author =

    A classification. Computational Statistics & Data Analysis , author =. 1992 , pages =

    work page 1992

  76. [77]

    Journal of Political Economy , author =

    The. Journal of Political Economy , author =. 1997 , pages =. doi:10.1086/262080 , language =

    work page doi:10.1086/262080 1997

  77. [78]

    Discrete

    Train, Kenneth , year =. Discrete

    work page

  78. [79]

    Journal of Classification , author =

    Large-sample results for optimization-based clustering methods , volume =. Journal of Classification , author =. 1991 , pages =. doi:10.1007/BF02616246 , abstract =

    work page doi:10.1007/bf02616246 1991

  79. [80]

    Biometrika , author =

    Asymptotic. Biometrika , author =. 1978 , note =. doi:10.2307/2335205 , abstract =

    work page doi:10.2307/2335205 1978

  80. [81]

    Statistics and Computing , author =

    An online classification. Statistics and Computing , author =. 2007 , pages =. doi:10.1007/s11222-007-9017-z , abstract =

    work page doi:10.1007/s11222-007-9017-z 2007

Showing first 80 references.