Pith. sign in

REVIEW 2 major objections 4 minor 58 references

SPOT-IC: Improving prediction for interval-censored data via survival probability transfer

T0 review · 2 major / 4 minor · reviewed 2026-07-13 · grok-4.5

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

Pith's one-line read Transferring survival probabilities—not model parameters—from arbitrary sources improves interval-censored prediction and can beat the target-only rate whenever at least one source is informative.

desk verdict Solid, usable transfer method for interval-censored prediction that actually relaxes the usual shared-model and shared-data requirements, with rate theory that holds under its stated premises. read the letter →

arxiv 2607.09640 v1 pith:NHQ6RDOP submitted 2026-07-10 stat.ME

Jingyi Hu , Yu Gu This is my paper
classification stat.ME MSC 62N0162N0262G2062F12
keywords intervalcensoringtransferlearningsurvivalprobabilitytransformationmodelsQ-aggregationEMalgorithmnegative
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

When disease onset is known only through wide exam intervals and short follow-up, standard survival prediction from a small target study is weak. This paper argues that you can still improve those predictions by borrowing survival-probability estimates from related source studies, even when the sources use different models and you never see their individual records. The method, SPOT-IC, adds a cross-entropy penalty that pulls the target survival surface toward the source probabilities and optimizes the result with a simple EM algorithm. With several sources of unknown quality it adaptively aggregates candidates so noninformative sources are downweighted. Theory and experiments claim a faster convergence rate than using the target data alone as soon as any source is accurate enough for the target population, including in an Alzheimer’s conversion application.

What carries the argument

The cross-entropy survival-probability penalty ψ_m(β,Λ), equivalent to a weighted current-status log-likelihood and therefore solvable by a Poisson-augmented EM algorithm that never needs raw source records; multi-source candidates are then combined by minimizing a Q-aggregation criterion with entropy regularization.

What would settle it

Simulate a target transformation model and sources whose survival curves differ by a fixed large gap on the target support; if, after optimal tuning and Q-aggregation, SPOT-IC’s L2 error and integrated Brier score still fail to beat the pure target-only estimator across replications, the central rate claim fails in that regime.

Watch

Extended reading notes

Core claim

For semiparametric transformation models with mixed-case interval-censored data, maximizing the target log-likelihood plus a survival-probability cross-entropy penalty—and, with multiple sources, Q-aggregating the resulting candidates—yields an estimator whose L2 survival error converges faster than the classical target-only n^{-2/3} rate whenever at least one source satisfies η_k + q_k = o(n^{-2/3}).

Load-bearing premise

The rate gain holds only if at least one source’s true survival function is close enough to the target’s on the times and covariates that matter; if every source is systematically off on that domain, the method need not beat target-only estimation.

Editorial extensions

If this is right

  • Small target cohorts with wide censoring intervals can improve prediction without sharing raw source records or matching source model structure.
  • Privacy-restricted biobanks and online risk calculators can serve as sources by sharing only survival-probability estimates.
  • Probability-level transfer can still help when sources differ in baseline hazard or model class (Cox, odds, AFT), provided survival curves remain close on the target domain.
  • Negative transfer can be controlled by data-adaptive Q-aggregation rather than by pre-selecting informative sources.
  • Transfer strength can be guided by the empirical distance between source and target-only survival estimates.

Reading between the lines

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

  • The same probability-transfer idea could calibrate large external AI survival predictors that are biased for a local clinic population.
  • Subject- or subgroup-specific aggregation weights (only sketched in the discussion) would better handle heterogeneous target cohorts.
  • Because only survival curves need to be shared, multi-site prediction becomes feasible without federated model training.
  • Comparing source versus target nonparametric survival curves is a practical pre-transfer screen for whether η_k is small enough to expect a rate gain.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

2 major / 4 minor

Summary. The paper proposes SPOT-IC, a transfer-learning procedure for prediction under mixed-case interval censoring. The target is modeled by a semiparametric transformation model (1); source studies contribute only precomputed survival-function estimators Š(t|X), which may come from arbitrary models and need not share covariates or individual-level data. Transfer is effected by a cross-entropy penalty ψ_m that is shown to be equivalent to a weighted current-status likelihood, yielding a stable EM algorithm via Poisson data augmentation. For multiple sources of unknown quality the authors form candidate estimators by single-source SPOT-IC plus a target-only estimator, then combine them by Q-aggregation (Algorithm 1). Theorems 1–3 give L2 rates: single-source candidates achieve O_p{n^{-2/3} ∧ (η_k + q_k)} and the aggregated estimator attains, up to an n^{-1} term, the rate of the best candidate, hence improves on the classical target-only n^{-2/3} rate whenever at least one source is sufficiently informative. Simulations (single- and multi-source, model mismatch, negative transfer) and an ADNI application support the claims.

Significance. The work addresses a genuine practical gap: interval-censored chronic-disease cohorts are often small and heavily censored, while external survival calculators or biobanks cannot share raw records. By transferring survival probabilities rather than parameters, SPOT-IC removes the usual requirements of identical model structure, common covariates, and individual-level source data. The multi-source Q-aggregation step supplies a theoretically justified safeguard against negative transfer. The rate improvement over the nonparametric MLE of Zeng et al. (2016) is non-trivial and is obtained under standard interval-censoring regularity plus a transparent similarity condition. The EM equivalence and the explicit oracle inequality for the aggregated estimator are concrete technical contributions that other transfer methods for censored data can build on.

major comments (2)
  1. Condition 7 (unique maximizer of the population penalized criterion) together with the requirement η_k + q_k = o(n^{-2/3}) is load-bearing for the rate gain claimed in Theorems 2–3. The paper correctly reverts to the target-only rate when the condition fails, but the manuscript never supplies a practical diagnostic or a finite-sample check that the selected ξ_n is near the theoretically optimal ξ_opt_n of Theorem 2. A short simulation or ADNI sensitivity plot that reports the realized ξ_n versus the plug-in ξ_opt_n, and the resulting L2 error, would make the central claim more credible to practitioners.
  2. In the multi-source theory (Theorem 3) the aggregation set I_agg is treated as independent of the screening estimators. Remark 1 mentions cross-fitting, yet the formal statement and the ADNI analysis appear to use a single split. Because the target sample is only n = 152 (ADNI) or 150 (simulations), the variability induced by the split can be large; the paper should either report cross-fitted results as the primary multi-source estimator or quantify the split-to-split variability of the aggregation weights and of IBS/NLL.
minor comments (4)
  1. Table 1 reports medians and MADs; the multi-source Figure 1 reports means ± SD. A uniform reporting convention would aid comparison.
  2. The construction of the artificial points (Ỹ, X̃) for the penalty is left somewhat informal (“uniform or truncated exponential”). A precise default recommendation would improve reproducibility.
  3. In the ADNI multi-source analysis the source models for TSDE/MAE are forced to be Cox while SPOT-IC is free to choose the transformation parameter; a brief note that this is required by those competitors (rather than an unfair advantage) would be helpful.
  4. A few typographical slips: “its jingyi@…” on the title page; occasional missing spaces around math operators; “Q-aggregation” sometimes written with an accented é and sometimes without.

Circularity Check

0 steps flagged · score 1.0 of 10

No significant circularity: rates are proved for a well-defined penalized M-estimator relative to external truth S_0 and external source rates q_k; self-citations supply tools, not the claim by construction.

full rationale

The derivation chain is standard transfer-learning M-estimation. The objective (2) is n^{-1} log L_n(β,Λ) + ξ_n ψ_m(β,Λ), with ψ_m a cross-entropy penalty that shrinks the target survival function toward an externally supplied source estimator Š; maximizers are characterized under Conditions 1–7, and Theorems 1–3 bound E∥Ŝ − S_0∥²_L2 in terms of the target sample size n, the source estimation rates q_k, and the population discrepancy η_k = E∥S_k − S_0∥²_L2. Those quantities are not fitted from the same functional being “predicted”: S_0 is the true target survival function, q_k is the rate of an arbitrary external source procedure, and ξ_n / ξ_θ are chosen by cross-validation rather than by equating the claim to a fitted constant. The multi-source step uses Q-aggregation (Lecué & Rigollet) on candidate estimators, with an oracle inequality that reverts to the target-only n^{-2/3} rate when no source is informative—so the rate gain is conditional, not forced. Self-citations (Gu et al. 2026 for the right-censored precursor; Zeng et al. for interval-censored NPMLE/EM tools) provide background machinery; Remark 3 explicitly develops new Hellinger/Fatou arguments for the interval-censored case rather than importing the target rate by definition. Condition 7 is an ordinary identifiability assumption on the population objective, not a uniqueness theorem that forbids alternatives by self-citation. Simulations and ADNI evaluate against held-out or true S_0. No step reduces Eq. X to Eq. Y by construction or renames a fit as a first-principles prediction.

Assumptions & free parameters 5 free parameters · 5 assumptions · 2 invented entities

The central rate claims rest on standard semiparametric interval-censoring regularity, a frailty-induced transformation model for the target, black-box source rate assumptions, an identifiability condition for the penalized limit, and free tuning parameters chosen by CV. No new physical entities; the main inventions are methodological (penalty, EM equivalence, aggregation).

free parameters (5)
  • ξ_n (single-source transfer strength)
    Controls how hard the target likelihood is pulled toward the source survival estimator; selected by cross-validation; optimal form given in Theorem 2 depends on estimated source–target discrepancy.
  • ξ_θ (Q-aggregation entropy weight)
    Stabilizes aggregation weights on the simplex; required ≥ C_0 theoretically and chosen by CV in practice.
  • Transformation frailty variance r (or G family)
    Target (and often source) logarithmic transformation parameter selected by AIC grid search; affects the working model though theory allows general G.
  • m and sampling law of (Ỹ, X̃) for the penalty
    Monte Carlo size and distribution used to form the empirical cross-entropy penalty; m→∞ in theory, practical choice affects finite-sample behavior.
  • Sample-split ratio |I_scr|/n and cross-fitting scheme
    Screening vs aggregation split (and optional multi-split averaging) is a design choice that affects multi-source performance when n is small.
assumptions (5)
  • domain assumption Mixed-case interval censoring with examination process independent of T given X; standard regularity on β0, Λ0, X, examination gaps and densities (Conditions 1–3).
    Standard for nonparametric MLE of interval-censored transformation models (Zeng et al.); required for likelihood validity and rates.
  • domain assumption Target follows a frailty-induced semiparametric transformation model with twice-differentiable G (Condition 4).
    Working model for the target estimator; sources need not share it.
  • domain assumption Each source estimator Ŝ_k converges in L2 at some rate q_k on the target covariate distribution (Condition 5).
    Black-box rate assumption; allows arbitrary source methods but the transfer gain depends on q_k + η_k.
  • ad hoc to paper Penalized population criterion has a unique maximizer (β*, Λ*) with smooth positive derivative Λ* (Condition 7).
    Identifiability of the transfer objective; needed for consistency when ξ is finite or infinite.
  • standard math Q-aggregation oracle inequality style bound under entropy penalty with ξ_θ ≥ C_0 (Theorem 3).
    Builds on Lecué & Rigollet / Dai et al. Q-aggregation theory with paper-specific arguments for estimation error under interval-censored likelihood.
invented entities (2)
  • SPOT-IC cross-entropy survival-probability penalty ψ_m and its Poisson/current-status EM equivalence
    purpose: Transfer source survival curves into the target interval-censored MLE without parameter matching or raw source data.
    Methodological construct defined in §2.2; not a physical entity; independent evidence is empirical/theoretical performance, not an external measurement.
  • Multi-source SPOT-IC Q-aggregation of candidate survival estimators
    purpose: Adaptively combine target-only and single-source transferred estimators to resist negative transfer.
    Adaptation of known Q-aggregation to interval-censored survival candidates; validated in simulations, not an external natural object.

how reviews work

0 comments
Cite this review

Pith. "Pith review of SPOT-IC: Improving prediction for interval-censored data via survival probability transfer." pith.science (2026). https://pith.science/paper/NHQ6RDOP

@misc{pith2026260709640,
  author       = {Pith},
  title        = {Pith review of: SPOT-IC: Improving prediction for interval-censored data via survival probability transfer},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/NHQ6RDOP}},
  note         = {Machine review of arXiv:2607.09640}
}
read the original abstract

Accurate prediction with interval-censored data is particularly challenging when censoring intervals are wide and follow-up is limited, as is common in studies of chronic diseases. Although auxiliary information from source studies may improve prediction in a target study, existing transfer learning methods typically impose restrictive assumptions on model or parameter similarity, or require access to individual-level source data. We propose a novel transfer learning method for interval-censored data that allows arbitrary source models and avoids sharing source data. Our approach transfers survival probability information from source studies through a carefully designed penalty and enables efficient computation via a simple EM algorithm. When multiple source studies are available and their informativeness is unknown, we further develop a data-adaptive aggregation procedure that is robust to negative transfer. Theoretical analysis shows that the proposed estimator attains a faster convergence rate than the target-only estimator whenever at least one source study is sufficiently informative. Extensive simulation studies and an application to data from the Alzheimer's Disease Neuroimaging Initiative demonstrate the effectiveness of our approach.

Figures

Figures reproduced from arXiv: 2607.09640 by the authors.

Figure 1
Figure 1. Prediction results in the multi-source simulation study under varying numbers of informative sources. Points [PITH_FULL_IMAGE:figures/full_fig_p012_1.png] view at source ↗
Figure 2
Figure 2. Turnbull estimators of the survival functions for the non-White (target) and White (source) cohorts [PITH_FULL_IMAGE:figures/full_fig_p014_2.png] view at source ↗

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

58 extracted references · 1 canonical work pages

  1. [1]

    Management Science , volume =

    Bastani, Hamsa , title =. Management Science , volume =. 2021 , doi =

    2021

  2. [2]

    Tony and Li, Hongzhe , title =

    Li, Sai and Cai, T. Tony and Li, Hongzhe , title =. Journal of the Royal Statistical Society Series B: Statistical Methodology , volume =

  3. [3]

    Tony and Li, Hongzhe , title =

    Li, Sai and Zhang, Linjun and Cai, T. Tony and Li, Hongzhe , title =. Journal of the American Statistical Association , volume =

  4. [4]

    Journal of the American Statistical Association , volume =

    Ye Tian and Yang Feng , title =. Journal of the American Statistical Association , volume =. 2023 , doi =

    2023

  5. [5]

    , title =

    Li, Yan and Wang, Lu and Wang, Jie and Ye, Jieping and Reddy, Chandan K. , title =. 2016 IEEE 16th International Conference on Data Mining (ICDM) , pages =. 2016 , publisher =

    2016

  6. [6]

    Journal of the American Statistical Association , volume =

    Li, Ziyi and Shen, Yu and Ning, Jing , title =. Journal of the American Statistical Association , volume =. 2023 , doi =

    2023

  7. [7]

    Statistical Analysis and Data Mining: The ASA Data Science Journal , volume =

    Xie, Mengqi and Hu, Tao and Zhou, Jie , title =. Statistical Analysis and Data Mining: The ASA Data Science Journal , volume =. 2024 , doi =

    2024

  8. [8]

    Biostatistics , volume =

    Lu, Yuying and Gu, Tian and Duan, Rui , title =. Biostatistics , volume =

Show all 58 references
  1. [9]

    Journal of Computational and Graphical Statistics , year =

    Yichen Lou and Mingyue Du and Jianguo Sun , title =. Journal of Computational and Graphical Statistics , year =

  2. [10]

    Gu, Yu and Zeng, Donglin and Lin, D. Y. , title =. 2026 , note =

    2026

  3. [11]

    Tony and Wei, Hongji , title =

    Cai, T. Tony and Wei, Hongji , title =. The Annals of Statistics , volume =

  4. [12]

    Tony and Kim, Dongwoo and Pu, Hongming , title =

    Cai, T. Tony and Kim, Dongwoo and Pu, Hongming , title =. The Annals of Statistics , volume =

  5. [13]

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

    Gu, Tian and Han, Yi and Duan, Rui , title =. Journal of the Royal Statistical Society Series B: Statistical Methodology , volume =. 2024 , month =

    2024

  6. [14]

    Optimal learning with Q-aggregation , volume=

    Lecué, Guillaume and Rigollet, Philippe , year=. Optimal learning with Q-aggregation , volume=. doi:10.1214/13-aos1190 , journal=

    doi

  7. [15]

    Zeng, Donglin and Mao, Lu and Lin, D. Y. , title =. Biometrika , volume =. 2016 , month =

    2016

  8. [16]

    Statistics in Medicine , year =

    Graf, Erika and Schmoor, Claudia and Sauerbrei, Willi and Schumacher, Martin , title =. Statistics in Medicine , year =

  9. [17]

    Statistics in Medicine , volume=

    On the C-statistics for evaluating overall adequacy of risk prediction procedures with censored survival data , author=. Statistics in Medicine , volume=. 2011 , publisher=

    2011

  10. [18]

    1996 , volume =

    Multivariable prognostic models: Issues in developing models, evaluating assumptions and adequacy, and measuring and reducing errors , journal =. 1996 , volume =

    1996

  11. [19]

    Yangjianchen Xu and Donglin Zeng and D. Y. Lin , title =. Journal of the American Statistical Association , pages =. 2025 , publisher =

    2025

  12. [20]

    van der Vaart, A. W. and Wellner, J. A. , title =

  13. [21]

    van de Geer, S. A. , title =

  14. [22]

    Zeng, Donglin and Gao, Fei and Lin, D. Y. , title =. Biometrika , volume =. 2017 , doi =

    2017

  15. [23]

    1991 , isbn =

    Ledoux, Michel and Talagrand, Michel , title =. 1991 , isbn =

    1991

  16. [24]

    The Annals of Statistics , volume=

    Efficient estimation for the proportional hazards model with interval censoring , author=. The Annals of Statistics , volume=. 1996 , publisher=

    1996

  17. [25]

    Journal of the American Statistical Association , volume=

    Sieve estimation for the proportional-odds failure-time regression model with interval censoring , author=. Journal of the American Statistical Association , volume=. 1997 , publisher=

    1997

  18. [26]

    The Annals of Statistics , number =

    Dong Dai and Philippe Rigollet and Tong Zhang , title =. The Annals of Statistics , number =. 2012 , doi =

    2012

  19. [27]

    1976 , publisher=

    Turnbull, Bruce W , journal=. 1976 , publisher=

    1976

  20. [28]

    Angelopoulos and Stephen Bates and Clara Fannjiang and Michael I

    Anastasios N. Angelopoulos and Stephen Bates and Clara Fannjiang and Michael I. Jordan and Tijana Zrnic , title =. Science , volume =. 2023 , doi =

    2023

  21. [29]

    Angelopoulos and John C

    Anastasios N. Angelopoulos and John C. Duchi and Tijana Zrnic , year=

  22. [30]

    2026 , note=

    Generalized Prediction-Powered Inference, with Application to Binary Classifier Evaluation , author=. 2026 , note=

    2026

  23. [31]

    Petersen, R. C. and Aisen, P. S. and Beckett, L. A. and Donohue, M. C. and Gamst, A. C. and Harvey, D. J. and Jack, C. R. and Jagust, W. J. and Shaw, L. M. and Toga, A. W. and Trojanowski, J. Q. and Weiner, M. W. , title =. Neurology , volume =

  24. [32]

    Angelopoulos, A. N. , Bates, S. , Fannjiang, C. , Jordan, M. I. & Zrnic, T. (2023). Prediction-powered inference. Science 382, 669--674

    2023

  25. [33]

    Angelopoulos, A. N. , Duchi, J. C. & Zrnic, T. (2024). PPI++ : Efficient prediction-powered inference. arXiv : 2311.01453

    2024 arXiv

  26. [34]

    Bastani, H. (2021). Predicting with proxies: Transfer learning in high dimension. Management Science 67, 2964--2984

    2021

  27. [35]

    Cai, T. T. , Kim, D. & Pu, H. (2024). Transfer learning for functional mean estimation: Phase transition and adaptive algorithms. The Annals of Statistics 52, 654--678

    2024

  28. [36]

    Cai, T. T. & Wei, H. (2021). Transfer learning for nonparametric classification: Minimax rate and adaptive classifier. The Annals of Statistics 49, 100--128

    2021

  29. [37]

    , Rigollet, P

    Dai, D. , Rigollet, P. & Zhang, T. (2012). Deviation optimal learning using greedy Q -aggregation . The Annals of Statistics 40, 1878 -- 1905

    2012

  30. [38]

    , Schmoor, C

    Graf, E. , Schmoor, C. , Sauerbrei, W. & Schumacher, M. (1999). Assessment and comparison of prognostic classification schemes for survival data. Statistics in Medicine 18, 2529--2545

    1999

  31. [39]

    , Han, Y

    Gu, T. , Han, Y. & Duan, R. (2024). Robust angle-based transfer learning in high dimensions. Journal of the Royal Statistical Society Series B: Statistical Methodology 87, 723--745

    2024

  32. [40]

    , Zeng, D

    Gu, Y. , Zeng, D. & Lin, D. Y. (2026). Prediction-oriented transfer learning for survival analysis. arXiv : 2603.11465

    2026

  33. [41]

    Huang, J. (1996). Efficient estimation for the proportional hazards model with interval censoring. The Annals of Statistics 24, 540--568

    1996

  34. [42]

    & Rossini, A

    Huang, J. & Rossini, A. (1997). Sieve estimation for the proportional-odds failure-time regression model with interval censoring. Journal of the American Statistical Association 92, 960--967

    1997

  35. [43]

    & Rigollet, P

    Lecué, G. & Rigollet, P. (2014). Optimal learning with q-aggregation. The Annals of Statistics 42, 211--224

    2014

  36. [44]

    , Cai, T

    Li, S. , Cai, T. T. & Li, H. (2022). Transfer learning for high-dimensional linear regression: Prediction, estimation and minimax optimality. Journal of the Royal Statistical Society Series B: Statistical Methodology 84, 149--173

    2022

  37. [45]

    , Zhang, L

    Li, S. , Zhang, L. , Cai, T. T. & Li, H. (2024). Estimation and inference for high-dimensional generalized linear models with knowledge transfer. Journal of the American Statistical Association 119, 1274--1285

    2024

  38. [46]

    , Wang, L

    Li, Y. , Wang, L. , Wang, J. , Ye, J. & Reddy, C. K. (2016). Transfer learning for survival analysis via efficient l2,1-norm regularized cox regression. In 2016 IEEE 16th International Conference on Data Mining (ICDM). IEEE

    2016

  39. [47]

    , Shen, Y

    Li, Z. , Shen, Y. & Ning, J. (2023). Accommodating time-varying heterogeneity in risk estimation under the cox model: A transfer learning approach. Journal of the American Statistical Association 118, 2276--2287

    2023

  40. [48]

    Lou, Y. , Du, M. & Sun, J. (2026). Two transfer learning approaches for regression analysis of high-dimensional interval-censored failure time data. Journal of Computational and Graphical Statistics , 1--8

    2026

  41. [49]

    Lu, Y. , Gu, T. & Duan, R. (2026). Adaptive transfer learning for time-to-event modeling with applications in disease risk assessment. Biostatistics 27, kxag011

    2026

  42. [50]

    Petersen, R. C. , Aisen, P. S. , Beckett, L. A. , Donohue, M. C. , Gamst, A. C. , Harvey, D. J. , Jack, C. R. , Jagust, W. J. , Shaw, L. M. , Toga, A. W. , Trojanowski, J. Q. & Weiner, M. W. (2010). Alzheimer's Disease Neuroimaging Initiative (ADNI) . Neurology 74, 201--209

    2010

  43. [51]

    & Feng, Y

    Tian, Y. & Feng, Y. (2023). Transfer learning under high-dimensional generalized linear models. Journal of the American Statistical Association 118, 2684--2697

    2023

  44. [52]

    Turnbull, B. W. (1976). The Empirical Distribution Function with Arbitrarily Grouped, Censored and Truncated Data . Journal of the Royal Statistical Society: Series B (Methodological) 38, 290--295

    1976

  45. [53]

    , Cai, T

    Uno, H. , Cai, T. , Pencina, M. J. , D'Agostino, R. B. & Wei, L.-J. (2011). On the c-statistics for evaluating overall adequacy of risk prediction procedures with censored survival data. Statistics in Medicine 30, 1105--1117

    2011

  46. [54]

    Xie, M. , Hu, T. & Zhou, J. (2024). Transfer learning under the cox model with interval-censored data. Statistical Analysis and Data Mining: The ASA Data Science Journal 17, e11680

    2024

  47. [55]

    , Zeng, D

    Xu, Y. , Zeng, D. & Lin, D. Y. (2025). Checking the cox proportional hazards model with interval-censored data. Journal of the American Statistical Association , 1--16

    2025

  48. [56]

    , Gao, F

    Zeng, D. , Gao, F. & Lin, D. Y. (2017). Maximum likelihood estimation for semiparametric regression models with multivariate interval-censored data. Biometrika 104, 505--525

    2017

  49. [57]

    , Mao, L

    Zeng, D. , Mao, L. & Lin, D. Y. (2016). Maximum likelihood estimation for semiparametric transformation models with interval-censored data. Biometrika 103, 253--271

    2016

  50. [58]

    , Witten, D

    Zou, R. , Witten, D. & Williamson, B. (2026). Generalized prediction-powered inference, with application to binary classifier evaluation. arXiv : 2602.10332

    2026

Pith tools

Reviewed July 13, 2026 · model on record in the stance chip above.