pith. sign in

arxiv: 2508.14936 · v3 · submitted 2025-08-19 · 🧬 q-bio.QM · cs.AI· cs.LG· stat.AP· stat.ML

Can synthetic data reproduce real-world findings in epidemiology? A replication study using adversarial random forests

Jan Kapar , Kathrin G\"unther , Lori Ann Vallis , Klaus Berger , Nadine Binder , Hermann Brenner , Stefanie Castell , Beate Fischer
show 22 more authors
Volker Harth Bernd Holleczek Timm Intemann Till Ittermann Andr\'e Karch Thomas Keil Lilian Krist Berit Lange Michael F. Leitzmann Katharina Nimptsch Nadia Obi Iris Pigeot Tobias Pischon Tamara Schikowski B\"orge Schmidt Carsten Oliver Schmidt Anja M. Sedlmair Justine Tanoey Harm Wienbergen Andreas Wienke Claudia Wigmann Marvin N. Wright
This is my paper

Pith reviewed 2026-05-18 22:18 UTC · model grok-4.3

classification 🧬 q-bio.QM cs.AIcs.LGstat.APstat.ML
keywords synthetic dataepidemiologyadversarial random forestsdata replicationprivacy preservationstatistical utilitytabular data generation
0
0 comments X p. Extension

The pith

Synthetic data from adversarial random forests reproduces findings from real epidemiological studies

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

This paper tests whether synthetic data generated by adversarial random forests can produce the same statistical results as real data in epidemiological research. The authors replicated analyses from six different studies covering areas such as blood pressure, heart attacks, and diabetes using data from major cohorts. Results matched closely between real and synthetic versions, including in challenging cases with limited samples relative to the number of variables. A sympathetic reader would care because this method could solve problems of data access and privacy while allowing continued scientific progress. The approach also shows advantages in efficiency and balance of utility versus privacy compared to other synthesis techniques.

Core claim

Adversarial random forests generate synthetic versions of epidemiological datasets that allow replication of descriptive and inferential statistical analyses from original publications, with results aligning consistently across the tested studies even when sample sizes are small relative to dimensionality.

What carries the argument

Adversarial random forests, which use an adversarial training process with random forests to create synthetic tabular data that maintains statistical properties of the original data.

If this is right

  • Analyses can be performed on synthetic data without compromising participant privacy in large cohort studies.
  • Findings from published research can be verified or extended using accessible synthetic datasets.
  • Data synthesis becomes more practical for non-experts due to the method's computational efficiency.
  • Quality improves with reduced data dimensionality, pointing to benefits of variable selection before synthesis.

Where Pith is reading between the lines

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

  • If the alignment holds for more complex causal models, synthetic data could support a wider range of research questions beyond simple associations.
  • Public release of synthetic cohort data might accelerate collaborative research while meeting ethical standards.
  • Comparisons suggest this method could be adapted for other tabular data domains with similar privacy needs.

Load-bearing premise

Alignment of results on the selected analyses from these six publications is enough to indicate that synthetic data can reproduce key findings in epidemiological research more generally.

What would settle it

Finding a new study or analysis type where the conclusions drawn from synthetic data differ substantially from those based on the original data would challenge the reliability claim.

Figures

Figures reproduced from arXiv: 2508.14936 by Andreas Wienke, Andr\'e Karch, Anja M. Sedlmair, Beate Fischer, Berit Lange, Bernd Holleczek, B\"orge Schmidt, Carsten Oliver Schmidt, Claudia Wigmann, Harm Wienbergen, Hermann Brenner, Iris Pigeot, Jan Kapar, Justine Tanoey, Katharina Nimptsch, Kathrin G\"unther, Klaus Berger, Lilian Krist, Lori Ann Vallis, Marvin N. Wright, Michael F. Leitzmann, Nadia Obi, Nadine Binder, Stefanie Castell, Tamara Schikowski, Thomas Keil, Till Ittermann, Timm Intemann, Tobias Pischon, Volker Harth.

Figure 1
Figure 1. Figure 1: Full dataset replication of Figure 2 in Schikowski et al. [19]: differences of mean blood pressure values (in mmHg) [PITH_FULL_IMAGE:figures/full_fig_p005_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Full dataset replication of Figure 5 in Fischer et al. [20]: subcutaneous and visceral abdominal adipose tissue thickness [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Full dataset replication of separate logistic regressions per variable, each adjusted for age, sex, country of birth, and years [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Full dataset replication of Figure 2 in Breau et al. [22]: calculated average valid wear time minutes per day spent in SED, [PITH_FULL_IMAGE:figures/full_fig_p008_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Full and task-specific dataset replication of multivariable linear regression, Table 3 in Berger et al. [23]: relationship [PITH_FULL_IMAGE:figures/full_fig_p009_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Full and task-specific dataset replication of univariable Cox regressions, Table 2 in Tanoey et al. [24]: type 1 diabetes [PITH_FULL_IMAGE:figures/full_fig_p010_6.png] view at source ↗
read the original abstract

Synthetic data holds substantial potential to address practical challenges in epidemiology due to restricted data access and privacy concerns. However, many current methods suffer from limited quality, high computational demands, and complexity for non-experts. Furthermore, common evaluation strategies for synthetic data often fail to directly reflect statistical utility and measure privacy risks sufficiently. Against this background, a critical underexplored question is whether synthetic data can reliably reproduce key findings from epidemiological research while preserving privacy. We propose adversarial random forests (ARF) as an efficient and convenient method for synthesizing tabular epidemiological data. To evaluate its performance, we replicated statistical analyses from six epidemiological publications covering blood pressure, anthropometry, myocardial infarction, accelerometry, loneliness, and diabetes, from the German National Cohort (NAKO Gesundheitsstudie), the Bremen STEMI Registry U45 Study, and the Guelph Family Health Study. We further assessed how dataset dimensionality and variable complexity affect the quality of synthetic data, and contextualized ARF's performance by comparison with commonly used tabular data synthesizers in terms of utility, privacy, generalisation, and runtime. Across all replicated studies, results on ARF-generated synthetic data consistently aligned with original findings. Even for datasets with relatively low sample size-to-dimensionality ratios, replication outcomes closely matched the original results across descriptive and inferential analyses. Reduced dimensionality and variable complexity further enhanced synthesis quality. ARF demonstrated favourable performance regarding utility, privacy preservation, and generalisation relative to other synthesizers and superior computational efficiency.

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 / 3 minor

Summary. The paper claims that adversarial random forests (ARF) provide an efficient method for generating synthetic tabular epidemiological data that preserves privacy while reproducing key findings from real-world studies. The authors replicate descriptive and inferential analyses from six publications using data from the German National Cohort (NAKO), Bremen STEMI Registry U45 Study, and Guelph Family Health Study. They report consistent alignment between original and ARF-synthetic results across these studies, even in low sample-size-to-dimensionality regimes, with reduced dimensionality improving quality, and ARF outperforming other synthesizers on utility, privacy, generalization, and runtime.

Significance. If the central claim holds, the work would be significant for epidemiology and synthetic data research by offering empirical evidence that a computationally efficient, accessible method can support real research questions under privacy constraints. The replication design using actual published analyses is a strength over purely metric-based evaluations, and the multi-cohort, multi-study scope plus baseline comparisons add practical value.

major comments (2)
  1. [Abstract and Results] Abstract and Results: The claims that results 'consistently aligned with original findings' and 'closely matched' lack any quantitative support such as effect-size differences, equivalence tests, or agreement statistics between original and synthetic outputs. Without these, the degree of fidelity cannot be assessed rigorously, especially for inferential statistics where small shifts may change conclusions.
  2. [Evaluation and Discussion] Evaluation and Discussion: The manuscript tests only the specific descriptive and inferential analyses pre-selected from the six publications. This does not establish that the synthetic data preserves the joint distributions, conditional dependencies, or tail behavior needed for untested epidemiological questions; additional held-out analyses or dependency checks would be required to support the broader generalization claim.
minor comments (3)
  1. [Methods] Methods: Include implementation details for ARF (hyperparameters, training procedure) and how post-hoc analysis choices from the original papers were made to support reproducibility and reduce selection concerns.
  2. [Results] Figures: Add error bars, confidence intervals, or quantitative difference metrics to plots comparing original versus synthetic results to make alignment visually and quantitatively clearer.
  3. [Introduction] Introduction: Provide a short description or key reference for adversarial random forests to aid readers unfamiliar with the technique.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive comments that highlight opportunities to strengthen the quantitative rigor and scope of our claims. We address each major comment below.

read point-by-point responses

  1. Referee: [Abstract and Results] Abstract and Results: The claims that results 'consistently aligned with original findings' and 'closely matched' lack any quantitative support such as effect-size differences, equivalence tests, or agreement statistics between original and synthetic outputs. Without these, the degree of fidelity cannot be assessed rigorously, especially for inferential statistics where small shifts may change conclusions.

    Authors: We agree that the manuscript would benefit from explicit quantitative comparisons. In the revised version we will add tables reporting relative differences (in percent) for all key descriptive statistics and effect estimates, absolute differences in p-values, and, where appropriate, equivalence testing bounds for the inferential results. These additions will allow readers to judge the practical magnitude of any discrepancies. revision: yes

  2. Referee: [Evaluation and Discussion] Evaluation and Discussion: The manuscript tests only the specific descriptive and inferential analyses pre-selected from the six publications. This does not establish that the synthetic data preserves the joint distributions, conditional dependencies, or tail behavior needed for untested epidemiological questions; additional held-out analyses or dependency checks would be required to support the broader generalization claim.

    Authors: The central objective of the study is to evaluate whether synthetic data can reproduce the specific published findings that motivated the original analyses, not to certify the data for arbitrary downstream questions. Because the six replication targets were chosen precisely because they represent the primary scientific conclusions drawn from each cohort, successful reproduction directly addresses the paper’s research question. We will nevertheless revise the Discussion to (i) explicitly delimit the scope of our claims to reproduction of reported findings and (ii) acknowledge that broader utility for novel analyses would require additional validation. If space allows, we will also report pairwise correlation matrices or mutual-information summaries as a supplementary check on dependency preservation. revision: partial

Circularity Check

0 steps flagged

No circularity: empirical replication against external benchmarks

full rationale

This is an empirical replication study that applies ARF to generate synthetic versions of real epidemiological datasets and directly compares descriptive and inferential results on the synthetic data against the published findings from six independent external studies. No equations, fitted parameters, or derivations are present that reduce to the paper's own inputs by construction. The evaluation relies on external original results as the benchmark, satisfying the criterion for self-contained evidence against external benchmarks. Any citations to the ARF method itself are not load-bearing for the replication claim and do not create a self-referential loop.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The central claim rests on the domain assumption that statistical utility in selected analyses transfers to general epidemiological utility, plus standard assumptions about synthetic data preserving joint distributions.

axioms (1)
  • domain assumption Synthetic data that matches selected descriptive and inferential statistics will support reliable epidemiological conclusions in general.
    Core premise linking replication success to broader utility; invoked in the evaluation strategy described in the abstract.

pith-pipeline@v0.9.0 · 5946 in / 1153 out tokens · 35259 ms · 2026-05-18T22:18:53.186845+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

43 extracted references · 43 canonical work pages

  1. [1]

    HIPAA privacy rule and public health; guidance from CDC and the U.S

    Centers for Disease Control and Prevention (U.S.). HIPAA privacy rule and public health; guidance from CDC and the U.S. Department of Health and Human Services, 2003. https://stacks.cdc.gov/view/cdc/12138 (30 July 2025, date last accessed)

    work page 2003

  2. [2]

    Regulation (EU) 2016/679 of the European Parliament and of the Council, 2016

    European Parliament and Council of the European Union. Regulation (EU) 2016/679 of the European Parliament and of the Council, 2016. https://eur-lex.europa.eu/eli/reg/2016/679/oj (30 July 2025, date last accessed)

    work page 2016

  3. [3]

    Generative deep learning

    David Foster. Generative deep learning. O’Reilly Media, Sebastopol, 2022

    work page 2022

  4. [4]

    OpenAI. ChatGPT. https://openai.com/chatgpt (30 July 2025, date last accessed)

    work page 2025

  5. [5]

    OpenAI. DALL·E. https://openai.com/dall-e (30 July 2025, date last accessed)

    work page 2025

  6. [6]

    Synthetic data in biomedicine via generative artificial intelligence

    Boris van Breugel, Tennison Liu, Dino Oglic, and Mihaela van der Schaar. Synthetic data in biomedicine via generative artificial intelligence. Nat Rev Bioeng, 2:991–1004, 2024

    work page 2024

  7. [7]

    Synthetic data—what, why and how? Royal Society,

    James Jordon, Lukasz Szpruch, Florimond Houssiau, Mirko Bottarelli, Giovanni Cherubin, Carsten Maple, Samuel N Cohen, and Adrian Weller. Synthetic data—what, why and how? Royal Society,

    work page

  8. [8]

    https://royalsociety.org/-/media/policy/projects/privacy-enhancing-technologies/ Synthetic_Data_Survey-24.pdf (30 July 2025, date last accessed)

    work page 2025

  9. [9]

    Sys- tematic review of generative adversarial networks (GANs) for medical image classification and segmentation

    Jiwoong J Jeong, Amara Tariq, Tobiloba Adejumo, Hari Trivedi, Judy W Gichoya, and Imon Banerjee. Sys- tematic review of generative adversarial networks (GANs) for medical image classification and segmentation. J Digit Imaging, 35(2):137–152, 2022

    work page 2022

  10. [10]

    Diffusion models in medical imaging: a comprehensive survey

    Amirhossein Kazerouni, Ehsan Khodapanah Aghdam, Moein Heidari, Reza Azad, Mohsen Fayyaz, Ilker Haci- haliloglu, and Dorit Merhof. Diffusion models in medical imaging: a comprehensive survey. Med Image Anal, 88:e102846, 2023

    work page 2023

  11. [11]

    Navigating tabular data synthesis research understanding user needs and tool capabilities

    Maria F Davila R, Sven Groen, Fabian Panse, and Wolfram Wingerath. Navigating tabular data synthesis research understanding user needs and tool capabilities. SIGMOD Rec, 53(4):18–35, 2025

    work page 2025

  12. [12]

    An evaluation of synthetic data generators implemented in the python library synthcity

    Emma F ¨ossing and J¨org Drechsler. An evaluation of synthetic data generators implemented in the python library synthcity. In Priv Stat Databases, volume 14915 of LNCS, pages 178–193, 2024

    work page 2024

  13. [13]

    A note on the evaluation of generative models

    L Theis, A van den Oord, and M Bethge. A note on the evaluation of generative models. ICLR, 2016

    work page 2016

  14. [14]

    Tabular data generation: can we fool XGBoost? NeurIPS 2022 First Table Representation Workshop, 2022

    EL Hacen Zein and Tanguy Urvoy. Tabular data generation: can we fool XGBoost? NeurIPS 2022 First Table Representation Workshop, 2022

    work page 2022

  15. [15]

    Synthetic data generation for a longitudinal cohort study—evaluation, method extension and reproduction of published data analysis results

    Lisa K ¨uhnel, Julian Schneider, Ines Perrar, Tim Adams, Sobhan Moazemi, Fabian Prasser, Ute N¨othlings, Holger Fr¨ohlich, and Juliane Fluck. Synthetic data generation for a longitudinal cohort study—evaluation, method extension and reproduction of published data analysis results. Sci Rep, 14(1):e14412, 2024. 12 Can synthetic data reproduce real-world fin...

    work page 2024

  16. [16]

    An evaluation of the replicability of analyses using synthetic health data

    Khaled El Emam, Lucy Mosquera, Xi Fang, and Alaa El-Hussuna. An evaluation of the replicability of analyses using synthetic health data. Sci Rep, 14(1):e6978, 2024

    work page 2024

  17. [17]

    Adversarial random forests for density estimation and generative modeling

    David S Watson, Kristin Blesch, Jan Kapar, and Marvin N Wright. Adversarial random forests for density estimation and generative modeling. In AISTATS, volume 206 of PMLR, pages 5357–5375, 2023

    work page 2023

  18. [18]

    Wright, David S

    Marvin N. Wright, David S. Watson, Kristin Blesch, and Jan Kapar. arf: adversarial random forests. CRAN,

    work page

  19. [19]

    https://CRAN.R-project.org/package=arf (30 July 2025, date last accessed)

    work page 2025

  20. [20]

    Synthcity: a benchmark framework for diverse use cases of tabular synthetic data

    Zhaozhi Qian, Rob Davis, and Mihaela van der Schaar. Synthcity: a benchmark framework for diverse use cases of tabular synthetic data. Adv Neural Inf Process Syst, 36:3173–3188, 2023

    work page 2023

  21. [21]

    Tamara Schikowski, Claudia Wigmann, Kateryna B Fuks, Sabine Schipf, Margit Heier, Hannelore Neuhauser, Giselle Sarganas, Wolfgang Ahrens, Heiko Becher, Klaus Berger, et al. Blutdruckmessung in der NAKO— methodische Unterschiede, Blutdruckverteilung und Bekanntheit der Hypertonie im Vergleich zu anderen bev¨olkerungsbezogenen Studien in Deutschland [Blood ...

    work page 2020

  22. [22]

    Anthropometrische Messungen in der NAKO Gesundheitsstudie—mehr als nur Gr ¨oße und Gewicht [Anthropometric measures in the German Na- tional Cohort—more than weight and height]

    Beate Fischer, Anja M Sedlmeier, Saskia Hartwig, Christopher L Schlett, Wolfgang Ahrens, Fabian Bamberg, Hansj¨org Baurecht, Heiko Becher, Klaus Berger, Hans Binder, et al. Anthropometrische Messungen in der NAKO Gesundheitsstudie—mehr als nur Gr ¨oße und Gewicht [Anthropometric measures in the German Na- tional Cohort—more than weight and height]. Bundes...

    work page 2020

  23. [23]

    Lifestyle and metabolic risk factors in patients with early-onset myocardial infarction: a case-control study

    Harm Wienbergen, Daniel Boakye, Kathrin G ¨unther, Johannes Schmucker, Luis Alberto Mata Mar ´ın, Hatim Kerniss, Rajini Nagrani, Luise Struß, Stephan R ¨uhle, Tina Retzlaff, et al. Lifestyle and metabolic risk factors in patients with early-onset myocardial infarction: a case-control study. Eur J Prev Cardiol, 29(16):2076–2087, 2022

    work page 2076

  24. [24]

    ActiGraph cutpoints impact physical activity and sedentary behavior outcomes in young children

    Becky Breau, Hannah J Coyle-Asbil, Jess Haines, David WL Ma, and Lori Ann Vallis. ActiGraph cutpoints impact physical activity and sedentary behavior outcomes in young children. J Meas Phys Behav , 5(2):85–96, 2022

    work page 2022

  25. [25]

    Klaus Berger, Steffi Riedel-Heller, Alexander Pabst, Marcella Rietschel, Dirk Richter, and NAKO-Konsortium. Einsamkeit w ¨ahrend der ersten Welle der SARS-CoV-2-Pandemie—Ergebnisse der NAKO-Gesundheitsstudie [Loneliness during the first wave of the SARS-CoV-2 pandemic—results of the German National Cohort (NAKO)]. Bundesgesundheitsblatt, Gesundheitsforsch...

    work page 2021

  26. [26]

    Birth order, caesarean section, or daycare attendance in relation to child-and adult-onset type 1 diabetes: results from the German National Cohort

    Justine Tanoey, Christina Baechle, Hermann Brenner, Andreas Deckert, Julia Fricke, Kathrin G ¨unther, Andr ´e Karch, Thomas Keil, Alexander Kluttig, Michael Leitzmann, et al. Birth order, caesarean section, or daycare attendance in relation to child-and adult-onset type 1 diabetes: results from the German National Cohort. Int J Environ Res Public Health, ...

    work page 2022

  27. [27]

    Framework and baseline examination of the German National Cohort (NAKO)

    Annette Peters and German National Cohort (NAKO) Consortium. Framework and baseline examination of the German National Cohort (NAKO). Eur J Epidemiol, 37(10):1107–1124, 2022

    work page 2022

  28. [28]

    Guelph Family Health Study: pilot study of a home- based obesity prevention intervention

    Jess Haines, Sabrina Douglas, Julia A Mirotta, Carley O’Kane, Rebecca Breau, Kathryn Walton, Owen Krystia, Elie Chamoun, Angela Annis, Gerarda A Darlington, et al. Guelph Family Health Study: pilot study of a home- based obesity prevention intervention. Can J Public Health, 109(4):549–560, 2018

    work page 2018

  29. [29]

    Deep neural networks and tabular data: a survey

    Vadim Borisov, Tobias Leemann, Kathrin Seßler, Johannes Haug, Martin Pawelczyk, and Gjergji Kasneci. Deep neural networks and tabular data: a survey. IEEE Trans Neural Netw Learn Syst, 35(6):7499–7519, 2022

    work page 2022

  30. [30]

    Generating synthetic data is complicated: know your data and know your generator

    Jonathan Latner, Marcel Neunhoeffer, and J ¨org Drechsler. Generating synthetic data is complicated: know your data and know your generator. In Priv Stat Databases, volume 14915 of LNCS, pages 115–128. Springer, 2024

    work page 2024

  31. [31]

    Ross Quinlan

    J. Ross Quinlan. Induction of decision trees. Mach Learn, 1:81–106, 1986

    work page 1986

  32. [32]

    Why do tree-based models still outperform deep learning on typical tabular data? Adv Neural Inf Process Syst, 35:507–520, 2022

    L ´eo Grinsztajn, Edouard Oyallon, and Ga¨el Varoquaux. Why do tree-based models still outperform deep learning on typical tabular data? Adv Neural Inf Process Syst, 35:507–520, 2022

    work page 2022

  33. [33]

    An empirical evaluation of easily implemented, nonparametric methods for generating synthetic datasets

    J ¨org Drechsler and Jerome P Reiter. An empirical evaluation of easily implemented, nonparametric methods for generating synthetic datasets. Comput Stat Data Anal, 55(12):3232–3243, 2011

    work page 2011

  34. [34]

    Random forests

    Leo Breiman. Random forests. Mach learn, 45:5–32, 2001

    work page 2001

  35. [35]

    Mixture of distributions

    Herbert Robbins. Mixture of distributions. Ann Math Stat, 19(3):360–369, 1948

    work page 1948

  36. [36]

    An introduction to the bootstrap

    Bradley Efron and Robert J Tibshirani. An introduction to the bootstrap. Chapman and Hall/CRC, New York, 1994. 13 Can synthetic data reproduce real-world findings in epidemiology?

    work page 1994

  37. [37]

    The PHQ-9: validity of a brief depression severity measure

    Kurt Kroenke, Robert L Spitzer, and Janet BW Williams. The PHQ-9: validity of a brief depression severity measure. J Gen Intern Med, 16(9):606–613, 2001

    work page 2001

  38. [38]

    A brief measure for assessing generalized anxiety disorder: the GAD-7

    Robert L Spitzer, Kurt Kroenke, Janet BW Williams, and Bernd L¨owe. A brief measure for assessing generalized anxiety disorder: the GAD-7. Arch Intern Med, 166(10):1092–1097, 2006

    work page 2006

  39. [39]

    Fairness without imputation: a decision tree approach for fair prediction with missing values

    Haewon Jeong, Hao Wang, and Flavio P Calmon. Fairness without imputation: a decision tree approach for fair prediction with missing values. Proc AAAI Conf Artif Intell, 36(9):9558–9566, 2022

    work page 2022

  40. [40]

    Missing value imputation with adversarial random forests—MissARF

    Pegah Golchian, Jan Kapar, David S Watson, and Marvin N Wright. Missing value imputation with adversarial random forests—MissARF. arXiv. doi:10.48550/arXiv.2507.15681, 21 July 2025, preprint: not peer reviewed

    work page doi:10.48550/arxiv.2507.15681 2025

  41. [41]

    Countarfactuals—generating plausible model-agnostic counterfactual explanations with adversarial ran- dom forests

    Susanne Dandl, Kristin Blesch, Timo Freiesleben, Gunnar K ¨onig, Jan Kapar, Bernd Bischl, and Marvin N Wright. Countarfactuals—generating plausible model-agnostic counterfactual explanations with adversarial ran- dom forests. In Explainable Artificial Intelligence. xAI 2024 , volume 2155 of CCIS, pages 85–107. Springer, 2024

    work page 2024

  42. [42]

    ML-Doctor: holistic risk assessment of inference attacks against machine learning models

    Yugeng Liu, Rui Wen, Xinlei He, Ahmed Salem, Zhikun Zhang, Michael Backes, Emiliano De Cristofaro, Mario Fritz, and Yang Zhang. ML-Doctor: holistic risk assessment of inference attacks against machine learning models. In Proc USENIX Secur Symp, pages 4525–4542, 2022. 14 Can synthetic data reproduce real-world findings in epidemiology? A Replication result...

    work page 2022

  43. [43]

    Median and percentile-based 95% confidence intervals of synthetic data results are printed in orange. PHQ-9, nine-item Patient Health Questionnaire; GAD-7, Generalized Anxiety Disorder seven-item scale; SD, standard deviation 37 Can synthetic data reproduce real-world findings in epidemiology? E.3 Full and task-specific dataset replication of multivariabl...

    work page 2020