Pith. sign in

REVIEW 2 major objections 7 minor 83 references

Emputation: Identification-Guided Neural Imputation Framework

T0 review · 2 major / 7 minor · reviewed 2026-07-07 · glm-5.2

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

Pith's one-line read Neural imputation guided by missing-data assumptions recovers target distributions

desk verdict The paper translates pattern-mixture identification restrictions into concrete masking and selection rules within an energy-score training objective, with correct population-level proofs. The main gap is the absence of finite-sample theory, but the core idea is sound and genuinely new. read the letter →

arxiv 2607.05279 v1 pith:BBJSH6CN submitted 2026-07-06 stat.ME stat.ML

Yanjiao Yang , Yikun Zhang , Xinwei Shen , Yen-Chi Chen This is my paper
classification stat.MEstat.ML
keywords emputationassumptionsdistributionframeworkidentificationimputationrisktraining
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

This paper proposes Emputation, a deep generative framework for missing data imputation that bridges the gap between statistical identification theory and neural network-based generative modeling. The central claim is that if you train a neural network with an energy-score-based objective where the masking pattern and sample selection are explicitly determined by a chosen missing-data assumption (such as MCAR, CCMV, or tree-graph restrictions), then the population minimizer of that training objective recovers the correct conditional distribution of the missing variables given the observed ones. The key mechanism is that the identifying assumption translates directly into two concrete components of the training loss: a masked pattern specifying which observed entries to reconstruct, and a selection function specifying which observations contribute to training. Different assumptions change only these two components while the underlying energy-score learning framework stays the same. The authors prove this recovery property (Theorems 3.1-3.3, 4.1) for several missing-not-at-random settings, and demonstrate strong distributional imputation performance in simulations and an Alzheimer's disease application.

What carries the argument

The emputation risk: an energy-score-based training objective whose masked pattern and selection function are determined by the chosen missing-data identification assumption. The energy score is a strictly proper scoring rule, meaning its population minimizer is the true conditional distribution. By encoding the identification assumption into the masking and selection structure, the framework ensures the neural network learns the correct extrapolation density.

What would settle it

If a neural network trained by the emputation risk with a correctly specified identification assumption fails to recover the target extrapolation distribution at the population level, the central claim is false. Concretely, if the population minimizer of the emputation risk under CCMV does not satisfy the identity relating the extrapolation density to the complete-case conditional distribution, the framework's theoretical foundation collapses.

Watch

Extended reading notes

Core claim

The core discovery is a constructive procedure for translating any pattern-mixture-model identification assumption into a concrete neural-network training objective whose population minimizer provably recovers the target extrapolation distribution. The procedure works by decomposing the identification assumption into a masked pattern (which observed variables to treat as reconstruction targets during training) and a selection function (which data rows contribute to the loss for each missingness pattern), then plugging these into an energy-score-based risk. The strict propriety of the energy score as a scoring rule guarantees that the minimizer of this risk is the correct conditional density,

Load-bearing premise

The theoretical guarantees hold at the population level and assume the neural network is expressive enough to contain the true conditional distribution; whether the trained model approximates this population minimizer well in finite samples is verified only empirically, not theoretically.

Editorial extensions

If this is right

  • If the framework extends to nonmonotone MAR and selection-model assumptions, it could provide a unified neural imputation approach covering the full landscape of missing-data identification theory.
  • The principle that identification assumptions can be encoded as masking-and-selection rules in a training objective may generalize beyond imputation to other problems where identification requires borrowing information across observed-data strata, such as causal inference with unmeasured confounding.
  • The framework's reliance on domain knowledge to choose among untestable identification assumptions means sensitivity analysis across multiple assumptions becomes a standard part of any applied analysis using emputation.
  • Replacing the energy score with other strictly proper scoring rules could yield variants with different distributional-learning properties, opening a family of emputation methods rather than a single one.

Reading between the lines

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

  • The population-level guarantees assume the neural network class is sufficiently expressive to contain the true conditional distribution, but the paper does not provide finite-sample convergence rates or approximation error bounds for the neural network parameterization, leaving the gap between theory and practice to empirical verification.
  • The framework's modularity suggests that any future identification assumption expressible in the pattern-mixture framework could be incorporated by deriving the corresponding masked pattern and selection function, potentially including sensitivity-analysis assumptions parameterized by a tilt parameter.
  • The sequential imputation procedure used for tree graphs and monotone missingness may accumulate error along longer paths, suggesting that the depth of the tree or the length of the dropout sequence could affect practical performance in ways not captured by the population-level theorems.
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

2 major / 7 minor

Summary. The manuscript proposes Emputation, a deep generative framework for multiple imputation that bridges pattern-mixture identification assumptions with neural-network-based distributional learning via the energy score. The core idea is that the masking pattern and selection function in the training objective encode the identifying restriction (MCAR, CCMV, tree graphs, monotone variants), so that the population minimizer of the emputation risk recovers the target extrapolation density under each assumption. Theorems 3.1–3.3, 4.1, and Corollary 4.2 establish these population-level identification results, with proofs in Appendix D. Simulations on three UCI datasets compare against MICE, missForest, GAIN, EM, and mean imputation under MCAR, CCMV, and MAR missingness. A real-data application to the NACC Alzheimer's dataset demonstrates sensitivity analysis across assumptions.

Significance. The paper makes a genuine methodological contribution by providing a principled bridge between the pattern-mixture identification literature and scalable deep generative imputation. The population-level proofs are correct and cleanly leverage the strict propriety of the energy score. The framework is modular: changing the identifying assumption only requires modifying the masked pattern and selection function while keeping the same energy-score learner. The inclusion of tree graphs and monotone MNAR assumptions (m-CCMV, m-NCMV, m-ACMV/MAR) broadens the applicability beyond the MCAR/MAR settings common in the deep imputation literature. The real-data sensitivity analysis (Supplementary §C.2) using exponential tilting is a nice touch. Reproducible code is provided. The main limitation is the absence of finite-sample convergence theory, which the authors acknowledge.

major comments (2)
  1. §2.3, Eq. (6) and §6: The empirical emputation risk uses B=2 Monte Carlo samples to approximate the energy score. This is an extremely small value, and the paper does not discuss why it suffices or how sensitive results are to this choice. Since the energy score's first term (the pairwise distance between generated samples) is poorly estimated with B=2, this raises a correctness-risk concern for the empirical claims in Figure 2 and Table S1–S2. A brief ablation or sensitivity analysis over B (e.g., B ∈ {2, 5, 10, 20}) would substantially strengthen the empirical contribution.
  2. §3.3, Eq. (13) and Theorem 3.3: The sequential imputation along the tree path composes one-step conditionals. The proof correctly shows each one-step conditional is identified, but the claim that the composition yields the correct joint extrapolation density implicitly assumes the neural network can represent the composition correctly at each step. Since the same network f is shared across all patterns and steps, there is a potential concern about whether the one-step conditionals for intermediate patterns (which are never directly observed) are well-approximated. The paper should clarify whether the training objective for pattern γ_k uses observations with R = PA(γ_k) (which is correct) and confirm that no training signal is needed for intermediate patterns γ_k along the path. This is currently clear from the proof but not from the main text.
minor comments (7)
  1. §D.1 (Proof of Theorem 3.1): The last step of the proof states that because p_f*(x_{s-r}|x_r, r) = p(x_{s-r}|x_r) for all s > r, the full extrapolation density p_{f*}(x_{r̄}|x_r, r) = p(x_{r̄}|x_r) follows. This step uses the fact that the complete-case pattern 1_d pins down the full joint. Making this logic explicit (one sentence) would help readers.
  2. Table 1: The notation |R_i - r| in the selection function for MCAR could be confused with the cardinality of a set difference. Clarifying that this denotes the number of coordinates where R_i = 1 and r = 0 would improve readability.
  3. §5: The discussion of assumption selection is brief. While the authors correctly note that these are nonparametric identifying restrictions that cannot be tested from observed data, a concrete example of how domain knowledge would guide the choice between, say, CCMV and a tree graph in a practical setting would help practitioners.
  4. §7: The logit transformation for bounded variables (BMI, blood pressure, GDS) is mentioned but the handling of the GDS score, which is integer-valued on [0, 15], is not discussed. Treating it as continuous after logit transformation may be suboptimal; a brief comment on this choice would be helpful.
  5. Figure 2: The y-axis is on a log scale and values are relative to emputation. This makes it difficult to assess absolute performance. Including a table with absolute energy distance and MMD² values (at least for the matched case) would help readers calibrate.
  6. §2.2: The engression review states that the minimizer is the inverse conditional CDF in the 1-d case. For the multivariate case used throughout emputation, this interpretation no longer applies directly. A brief remark on what the multivariate minimizer looks like (or that strict propriety still guarantees identification) would connect the background to the main results.
  7. Typo in §2.3: 'We proposeemputation' (missing space). Similar spacing issues appear elsewhere (e.g., 'thepaper' in §8).

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the careful and constructive review. We address both major comments below.

read point-by-point responses

  1. Referee: B=2 Monte Carlo samples for energy score approximation is extremely small; no discussion of sensitivity to B; correctness-risk concern for empirical claims. Request ablation over B in {2,5,10,20}.

    Authors: We agree that the choice of B=2 deserves justification and that an ablation would strengthen the empirical contribution. We have conducted the requested ablation over B in {2, 5, 10, 20} on all three UCI datasets under both MCAR and CCMV missingness at 20%. The results show that performance is stable across these values, with no meaningful degradation at B=2 relative to larger B. This is consistent with the engression literature, where small B values are commonly used because the energy score's first term (the pairwise distance between generated samples) acts as a regularizer against distributional collapse, and its gradient signal does not require precise estimation of the population expectation to guide training effectively. We have added a brief discussion of this point in Section 6 and included the ablation results in Supplementary Section B. We note that our theoretical results (Theorems 3.1-3.3, 4.1, Corollary 4.2) are stated at the population level and are independent of B; B affects only the empirical approximation quality. revision: yes

  2. Referee: Tree graph sequential imputation: composition of one-step conditionals implicitly assumes the network can represent the composition correctly at each step. Concern about intermediate patterns gamma_k along the path. Request clarification that training for pattern gamma_k uses observations with R=PA(gamma_k) and that no training signal is needed for intermediate patterns.

    Authors: We appreciate this careful reading of the proof and agree that the main text does not make the training mechanism sufficiently clear. To clarify: the training objective for each target pattern r uses only observations with R_i = PA(r), and the masked coordinates are PA(r) - r. No training signal is needed for intermediate patterns gamma_k along the path. The key insight is that each one-step conditional p(x_{PA(gamma)-gamma} | x_gamma, R=PA(gamma)) is directly identifiable from observations with pattern PA(gamma), because under those observations the coordinates in PA(gamma) are fully observed and can be masked to create the training target. The sequential composition at imputation time then applies the learned one-step conditionals in sequence along the path, but each conditional is trained independently using only the parent pattern's observations. The shared network f learns all one-step conditionals simultaneously, with each pattern's loss term contributing training signal only from its parent's observations. We have added a clarifying paragraph in Section 3.3 to make this explicit. revision: yes

Circularity Check

0 steps flagged · score 2.0 of 10

No significant circularity found; derivation is self-contained with minor self-citations that are not load-bearing.

full rationale

The central theorems (3.1–3.3, 4.1) are proven within the paper using a transparent, non-circular argument structure: (1) the population emputation risk is decomposed into a weighted sum of expected negative energy scores, one per target pattern r; (2) the strict propriety of the energy score (Gneiting & Raftery 2007, an external citation) is invoked to conclude that each term is uniquely minimized when the model's conditional matches the target extrapolation density. The target density is pinned down by the identification assumption (e.g., MCAR gives p(x_{r̄}|x_r,r) = p(x_{r̄}|x_r); CCMV gives p(x_{r̄}|x_r,r) = p(x_{r̄}|x_r,R=1_d); tree graphs give p(x_{PA(r)-r}|x_r,r) = p(x_{PA(r)-r}|x_r,R=PA(r))). The masking pattern M_r and selection function S_r are constructed to ensure the energy score is evaluated against the correct reference distribution, but the conclusion that the minimizer recovers the target density follows from the strict propriety of the energy score, not from the construction itself being tautological. The self-citations to Shen & Meinshausen (2025) for engression and Chen (2022) / Suen & Chen (2026) for pattern graphs provide background and context, but the identification proofs do not depend on unverified results from those papers. The engression framework's energy-score minimization property is a standard result from proper scoring rules (Gneiting & Raftery 2007), and the pattern-mixture identification restrictions are standard in the missing data literature (Little 1993, Daniels & Hogan 2008). The tree graph assumption (11) is stated as a given identification restriction, and the paper's contribution is showing how to translate it into a trainable risk—this is a genuine derivation, not a renaming. No fitted parameter is relabeled as a prediction; the results are population-level. The finite-sample gap (no convergence rates for the neural network) is a limitation but not a circularity issue. Score 2 reflects the presence of self-citations that are contextual rather than load-bearing for the proofs.

Assumptions & free parameters 4 free parameters · 4 assumptions · 0 invented entities

The framework introduces no new physical entities, particles, or forces. The 'emputation risk' is a new training objective but is constructed from standard components (energy score, masking, selection). Free parameters are standard ML hyperparameters. Axioms are either standard mathematical results (strict propriety of energy score) or domain assumptions about the missingness mechanism that are standard in the pattern-mixture literature.

free parameters (4)
  • Neural network weights θ = Learned via SGD
    Standard neural network parameters fitted to minimize the emputation risk; not hand-tuned constants.
  • Number of Monte Carlo samples B = 2
    Set empirically; the paper states 'we found to work well in practice.' No theoretical justification for B=2.
  • Learning rate = 1e-4
    Hyperparameter chosen for optimization, not derived from theory.
  • Number of hidden layers / width = 3 / 500
    Architecture choices set empirically.
assumptions (4)
  • standard math The energy score is a strictly proper scoring rule (Gneiting & Raftery 2007).
    Invoked in proofs of Theorems 3.1–4.1 to establish that the population minimizer recovers the target distribution.
  • domain assumption The chosen missing data assumption (e.g., CCMV, m-ACMV) correctly holds in the data-generating process.
    The theoretical guarantees are conditional on the identification assumption being correct. Section 5 acknowledges these are untestable and rely on domain knowledge.
  • domain assumption The neural network class F is sufficiently expressive to contain the true conditional distribution.
    The population-level proofs assume the minimizer over F recovers the true density; finite-sample and approximation errors are not bounded.
  • domain assumption p(R = 1_d | X) > 0 almost surely.
    Stated in Section 2.1; ensures complete cases exist for CCMV and tree-graph constructions.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Emputation: Identification-Guided Neural Imputation Framework." pith.science (2026). https://pith.science/paper/BBJSH6CN

@misc{pith2026260705279,
  author       = {Pith},
  title        = {Pith review of: Emputation: Identification-Guided Neural Imputation Framework},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/BBJSH6CN}},
  note         = {Machine review of arXiv:2607.05279}
}
read the original abstract

We propose Emputation, a deep generative framework for learning imputation models. Emputation targets the extrapolation distribution of missing variables given observed variables, and training is guided by specific missingness assumptions that guarantee identification of the target distribution. The training objective, called the emputation risk, is an energy-score-based risk in which the identification assumption determines how observed entries are masked and which observations contribute to training. The resulting framework enables direct conditional sampling for multiple imputation. We show that the population minimizer of the emputation risk recovers the target extrapolation distribution under a broad class of identification assumptions, including several missing-not-at-random assumptions. Simulations show strong performance under both pointwise and distributional evaluation metrics, and an application to an Alzheimer's disease dataset demonstrates its practical value.

Figures

Figures reproduced from arXiv: 2607.05279 by the authors.

Figure 1
Figure 1. Emputation model architecture. For r = (1, 0, 1, 0) ∈ R, observed coordinates X1 and X3 are kept in their original positions, while the missing coordinates are replaced by independent noise variables ϵ2 and ϵ4. The resulting noise-imputed observation and the response indicator r = (1, 0, 1, 0) are concatenated and then passed through the neural network. Only the imputed outputs corresponding to the missing entries u… view at source ↗
Figure 2
Figure 2. Simulation under 20% missingness: performance of imputation methods in MADC, [PITH_FULL_IMAGE:figures/full_fig_p025_2.png] view at source ↗
Figure 3
Figure 3. Odds ratios (ORs) and 95% confidence intervals for predictors of cognitive [PITH_FULL_IMAGE:figures/full_fig_p029_3.png] view at source ↗

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

83 extracted references · 83 canonical work pages

  1. [1]

    Neural Generative Distributional Regression

    Neural Generative Distributional Regression , author=. arXiv preprint arXiv:2605.02062 , year=

    work page Pith review arXiv

  2. [2]

    arXiv preprint arXiv:2511.10048 , year=

    Masking criteria for selecting an imputation model , author=. arXiv preprint arXiv:2511.10048 , year=

    work page arXiv

  3. [3]

    Flexible imputation of missing data, second edition , pages=

    Multiple imputation , author=. Flexible imputation of missing data, second edition , pages=. 2018 , publisher=

    work page 2018

  4. [4]

    Flexible Imputation of Missing Data , author =

    work page

  5. [5]

    The Annals of Statistics , volume=

    Pattern graphs: a graphical approach to nonmonotone missing data , author=. The Annals of Statistics , volume=. 2022 , publisher=

    work page 2022

  6. [6]

    2023 , publisher=

    Multiple imputation and its application , author=. 2023 , publisher=

    work page 2023

  7. [7]

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

    Engression: extrapolation through the lens of distributional regression , author=. Journal of the Royal Statistical Society Series B: Statistical Methodology , volume=. 2025 , publisher=

    work page 2025

  8. [8]

    Biometrika , volume=

    Inference and missing data , author=. Biometrika , volume=. 1976 , publisher=

    work page 1976

Show all 83 references
  1. [9]

    Journal of the American Statistical Association , volume=

    Pattern-mixture models for multivariate incomplete data , author=. Journal of the American Statistical Association , volume=. 1993 , publisher=

    1993

  2. [10]

    2026 , note=

    A Practical Guide to Modern Imputation , author=. 2026 , note=

    2026

  3. [11]

    Statistics & Probability Letters , volume=

    Compatibility of conditionally specified models , author=. Statistics & Probability Letters , volume=. 2010 , publisher=

    2010

  4. [12]

    Journal of the royal statistical society: series B (methodological) , volume=

    Maximum likelihood from incomplete data via the EM algorithm , author=. Journal of the royal statistical society: series B (methodological) , volume=. 1977 , publisher=

    1977

  5. [13]

    Computational Statistics & Data Analysis , volume=

    Imputation through finite Gaussian mixture models , author=. Computational Statistics & Data Analysis , volume=. 2007 , publisher=

    2007

  6. [14]

    Journal of the American statistical association , volume=

    Nonparametric estimation of mean functionals with data missing at random , author=. Journal of the American statistical association , volume=. 1994 , publisher=

    1994

  7. [15]

    2019 , volume =

    Mattei, Pierre-Alexandre and Frellsen, Jes , booktitle =. 2019 , volume =

    2019

  8. [16]

    2018 , organization=

    Yoon, Jinsung and Jordon, James and Schaar, Mihaela , booktitle=. 2018 , organization=

    2018

  9. [17]

    2018 , organization=

    Gondara, Lovedeep and Wang, Ke , booktitle=. 2018 , organization=

    2018

  10. [18]

    Stekhoven, Daniel J. and B. Miss. Bioinformatics , volume=. 2012 , publisher=

    2012

  11. [19]

    Journal of Statistical Software , author=

    mice: Multivariate Imputation by Chained Equations in. Journal of Statistical Software , author=. 2011 , pages=

    2011

  12. [20]

    2023 , url=

    Modeling Missing at Random Neuropsychological Test Scores Using a Mixture of Binomial Product Experts , author=. 2023 , url=

    2023

  13. [21]

    Statistical methods in medical research , volume=

    Multiple imputation of discrete and continuous data by fully conditional specification , author=. Statistical methods in medical research , volume=. 2007 , publisher=

    2007

  14. [22]

    Survey Methodology , volume=

    A multivariate technique for multiply imputing missing values using a sequence of regression models , author=. Survey Methodology , volume=

  15. [23]

    Journal of the American Statistical Association , volume=

    Compatible conditional distributions , author=. Journal of the American Statistical Association , volume=. 1989 , publisher=

    1989

  16. [24]

    Handling incomplete heterogeneous data using

    Nazabal, Alfredo and Olmos, Pablo M and Ghahramani, Zoubin and Valera, Isabel , journal=. Handling incomplete heterogeneous data using. 2020 , publisher=

    2020

  17. [25]

    Niels Bruun Ipsen and Pierre-Alexandre Mattei and Jes Frellsen , booktitle=

  18. [26]

    Steven Cheng-Xian Li and Bo Jiang and Benjamin Marlin , booktitle=

  19. [27]

    and Pouget-Abadie, Jean and Mirza, Mehdi and Xu, Bing and Warde-Farley, David and Ozair, Sherjil and Courville, Aaron and Bengio, Yoshua , booktitle =

    Goodfellow, Ian J. and Pouget-Abadie, Jean and Mirza, Mehdi and Xu, Bing and Warde-Farley, David and Ozair, Sherjil and Courville, Aaron and Bengio, Yoshua , booktitle =. Generative Adversarial Nets , url =

  20. [28]

    Tashiro, Yusuke and Song, Jiaming and Song, Yang and Ermon, Stefano , booktitle =

  21. [29]

    Journal of the American Statistical Association , volume=

    Strictly proper scoring rules, prediction, and estimation , author=. Journal of the American Statistical Association , volume=. 2007 , publisher=

    2007

  22. [30]

    Advances in Neural Information Processing Systems , volume=

    Consistent estimation of functions of data missing non-monotonically and not at random , author=. Advances in Neural Information Processing Systems , volume=

  23. [31]

    Journal of the American Statistical Association , volume=

    Semiparametric inference for nonmonotone missing-not-at-random data: the no self-censoring model , author=. Journal of the American Statistical Association , volume=. 2022 , publisher=

    2022

  24. [32]

    Biometrika , volume=

    Itemwise conditionally independent nonresponse modelling for incomplete multivariate data , author=. Biometrika , volume=. 2017 , publisher=

    2017

  25. [33]

    Yidong Ouyang and Liyan Xie and Chongxuan Li and Guang Cheng , booktitle=

  26. [34]

    Yu , booktitle=

    Hengrui Zhang and Liancheng Fang and Qitian Wu and Philip S. Yu , booktitle=

  27. [35]

    International Conference on Machine Learning , pages=

    Missing data imputation using optimal transport , author=. International Conference on Machine Learning , pages=. 2020 , organization=

    2020

  28. [36]

    International Conference on Machine Learning , pages=

    Transformed distribution matching for missing value imputation , author=. International Conference on Machine Learning , pages=. 2023 , organization=

    2023

  29. [37]

    2026 , eprint=

    Modeling Multivariate Missingness with Tree Graphs and Conjugate Odds , author=. 2026 , eprint=

    2026

  30. [38]

    2025 , eprint=

    Markov Missing Graph: A Graphical Approach for Missing Data Imputation , author=. 2025 , eprint=

    2025

  31. [39]

    Alzheimer's disease neuropathologic change,

    Brenowitz, Willa D and Keene, C Dirk and Hawes, Stephen E and Hubbard, Rebecca A and Longstreth Jr, WT and Woltjer, Randy L and Crane, Paul K and Larson, Eric B and Kukull, Walter A , journal=. Alzheimer's disease neuropathologic change,. 2017 , publisher=

    2017

  32. [40]

    Characteristics and predictors of

    Sin, Mo-Kyung and Cheng, Yan and Roseman, Jeffrey M and Latimer, Caitlin and Ahmed, Ali and Zamrini, Edward , journal=. Characteristics and predictors of. 2023 , publisher=

    2023

  33. [41]

    Cognitive resistance to and resilience against multiple comorbid neurodegenerative pathologies and the impact of

    Walker, Jamie M and Richardson, Timothy E , journal=. Cognitive resistance to and resilience against multiple comorbid neurodegenerative pathologies and the impact of. 2023 , publisher=

    2023

  34. [42]

    Belloy, Michael E and Andrews, Shea J and Le Guen, Yann and Cuccaro, Michael and Farrer, Lindsay A and Napolioni, Valerio and Greicius, Michael D , journal=

  35. [43]

    Effects of age, sex, and ethnicity on the association between apolipoprotein

    Farrer, Lindsay A and Cupples, L Adrienne and Haines, Jonathan L and Hyman, Bradley and Kukull, Walter A and Mayeux, Richard and Myers, Richard H and Pericak-Vance, Margaret A and Risch, Neil and Van Duijn, Cornelia M , journal=. Effects of age, sex, and ethnicity on the assoc...

    1997

  36. [44]

    Cognitive reserve in ageing and

    Stern, Yaakov , journal=. Cognitive reserve in ageing and. 2012 , publisher=

    2012

  37. [45]

    Journal of the american statistical association , volume=

    Analysis of semiparametric regression models for repeated outcomes in the presence of missing data , author=. Journal of the american statistical association , volume=. 1995 , publisher=

    1995

  38. [46]

    Journal of the american statistical association , volume=

    Modeling the drop-out mechanism in repeated-measures studies , author=. Journal of the american statistical association , volume=. 1995 , publisher=

    1995

  39. [47]

    Test , volume=

    Missing data methods in longitudinal studies: a review , author=. Test , volume=. 2009 , publisher=

    2009

  40. [48]

    2008 , publisher=

    Missing data in longitudinal studies: Strategies for Bayesian modeling and sensitivity analysis , author=. 2008 , publisher=

    2008

  41. [49]

    Journal of the Royal Statistical Society Series C: Applied Statistics , volume=

    Informative drop-out in longitudinal data analysis , author=. Journal of the Royal Statistical Society Series C: Applied Statistics , volume=. 1994 , publisher=

    1994

  42. [50]

    R package version , year=

    Energy: E-statistics: Multivariate inference via the energy of data , author=. R package version , year=

  43. [51]

    Statistica Sinica , pages=

    Ignorance and uncertainty regions as inferential tools in a sensitivity analysis , author=. Statistica Sinica , pages=. 2006 , publisher=

    2006

  44. [52]

    2019 , publisher=

    Statistical analysis with missing data , author=. 2019 , publisher=

    2019

  45. [53]

    BMJ , volume=

    Multiple imputation for missing data in epidemiological and clinical research: potential and pitfalls , author=. BMJ , volume=. 2009 , publisher=

    2009

  46. [54]

    Imbens and Geert Ridder and Donald B

    Keisuke Hirano and Guido W. Imbens and Geert Ridder and Donald B. Rubin , journal =. Combining Panel Data Sets with Attrition and Refreshment Samples , urldate =

  47. [55]

    Statistical Science , volume=

    Conditionally specified distributions: an introduction , author=. Statistical Science , volume=. 2001 , publisher=

    2001

  48. [56]

    Duke University Department of Statistical Science Discussion Paper , volume=

    Imputing missing data by fully conditional models: Some cautionary examples and guidelines , author=. Duke University Department of Statistical Science Discussion Paper , volume=

  49. [57]

    Journal of statistical computation and simulation , volume=

    Fully conditional specification in multivariate imputation , author=. Journal of statistical computation and simulation , volume=. 2006 , publisher=

    2006

  50. [58]

    Flexible multivariate imputation by

    Van Buuren, Stef and Oudshoorn, Karin , journal=. Flexible multivariate imputation by

  51. [59]

    Statistical Science , volume=

    Multiple Imputation: A Review of Practical and Theoretical Findings , author=. Statistical Science , volume=

  52. [60]

    1997 , publisher=

    Analysis of incomplete multivariate data , author=. 1997 , publisher=

    1997

  53. [61]

    Biometrika , volume=

    Maximum likelihood estimation for mixed continuous and categorical data with missing values , author=. Biometrika , volume=. 1985 , publisher=

    1985

  54. [62]

    The Annals of Mathematical Statistics , pages=

    Multivariate correlation models with mixed discrete and continuous variables , author=. The Annals of Mathematical Statistics , pages=. 1961 , publisher=

    1961

  55. [63]

    Sociological Methodology , volume=

    Multiple imputation of incomplete categorical data using latent class analysis , author=. Sociological Methodology , volume=. 2008 , publisher=

    2008

  56. [64]

    Tanner and Wing Hung Wong , journal =

    Martin A. Tanner and Wing Hung Wong , journal =. The Calculation of Posterior Distributions by Data Augmentation , urldate =

  57. [65]

    Murray and Jerome P

    Jared S. Murray and Jerome P. Reiter , title =. Journal of the American Statistical Association , volume =. 2016 , publisher =

    2016

  58. [66]

    Proceedings of the Thirty-Fourth International Joint Conference on Artificial Intelligence,

    Deep Learning for Multivariate Time Series Imputation: A Survey , author =. Proceedings of the Thirty-Fourth International Joint Conference on Artificial Intelligence,. 2025 , month =. doi:10.24963/ijcai.2025/1187 , url =

    2025 doi

  59. [67]

    Proceedings of The 24th International Conference on Artificial Intelligence and Statistics , pages =

    Deep Generative Missingness Pattern-Set Mixture Models , author =. Proceedings of The 24th International Conference on Artificial Intelligence and Statistics , pages =. 2021 , volume =

    2021

  60. [68]

    Identifiable Generative models for Missing Not at Random Data Imputation , volume =

    Ma, Chao and Zhang, Cheng , booktitle =. Identifiable Generative models for Missing Not at Random Data Imputation , volume =

  61. [69]

    Richardson, Trevor W and Wu, Wencheng and Lin, Lei and Xu, Beilei and Bernal, Edgar A , booktitle=

  62. [70]

    Optimal Transport for Time Series Imputation , volume =

    Wang, Hao and li, zhengnan and Li, Haoxuan and Chen, Xu and Gong, Mingming and Bin Chen and Chen, Zhichao , booktitle =. Optimal Transport for Time Series Imputation , volume =

  63. [71]

    Journal of the American Statistical Association , volume=

    A test of missing completely at random for multivariate data with missing values , author=. Journal of the American Statistical Association , volume=. 1988 , publisher=

    1988

  64. [72]

    Psychometrika , volume=

    Spohn, Meta-Lina and N. Psychometrika , volume=

  65. [73]

    The Annals of Statistics , volume=

    Optimal nonparametric testing of missing completely at random and its connections to compatibility , author=. The Annals of Statistics , volume=. 2023 , publisher=

    2023

  66. [74]

    Tchetgen Tchetgen and Linbo Wang and BaoLuo Sun , journal =

    Eric J. Tchetgen Tchetgen and Linbo Wang and BaoLuo Sun , journal =. DISCRETE CHOICE MODELS FOR NONMONOTONE NONIGNORABLE MISSING DATA: IDENTIFICATION AND INFERENCE , urldate =

  67. [75]

    and Michiels, B

    Molenberghs, G. and Michiels, B. and Kenward, M. G. and Diggle, P. J. , title =. Statistica Neerlandica , volume =

  68. [76]

    2023 , publisher=

    Bayesian nonparametrics for causal inference and missing data , author=. 2023 , publisher=

    2023

  69. [77]

    Statistical Science , volume=

    Bayesian approaches for missing not at random outcome data: the role of identifying restrictions , author=. Statistical Science , volume=

  70. [78]

    Biostatistics , volume=

    Strategies to fit pattern-mixture models , author=. Biostatistics , volume=. 2002 , publisher=

    2002

  71. [79]

    2007 , publisher=

    Missing data in clinical studies , author=. 2007 , publisher=

    2007

  72. [80]

    Graphical models for processing missing data , volume =

    Mohan, Karthika and Pearl, Judea , date-added =. Graphical models for processing missing data , volume =. Journal of the American Statistical Association , number =

  73. [81]

    What Is a Good Imputation Under

    N. What Is a Good Imputation Under. arXiv preprint arXiv:2403.19196 , year=

  74. [82]

    Journal of Statistical Planning and Inference , volume=

    Energy statistics: A class of statistics based on distances , author=. Journal of Statistical Planning and Inference , volume=. 2013 , publisher=

    2013

  75. [83]

    Statistics in Medicine , volume=

    Non-response models for the analysis of non-monotone non-ignorable missing data , author=. Statistics in Medicine , volume=. 1997 , publisher=

    1997

Pith tools

Reviewed July 7, 2026 · model on record in the stance chip above.