Pith. sign in

REVIEW 3 major objections 4 minor 64 references

SPYCE is a doubly robust estimator that recovers reliable rates of outcome change before Stage 1 in Huntington disease, even when study dropout depends on the outcome.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review · deepseek-v4-flash

2026-08-01 15:12 UTC pith:OWUSNYOX

load-bearing objection Real contribution — new doubly robust estimator plus efficient score for outcome-dependent censoring, with detailed proofs — but the guarantee is narrower than the banner: it all sits on C⊥X|Y,Z, and the automatic-efficiency nonparametric result is what I want verified. the 3 major comments →

arxiv 2607.18501 v1 pith:OWUSNYOX submitted 2026-07-20 stat.ME stat.AP

SPYCE: A Doubly Robust Estimator for Trials Targeting Early Huntington Disease under Outcome-Dependent Censoring

Kihyun Han , Yanyuan Ma , Karen Marder , Tanya P. Garcia This is my paper
classification stat.ME stat.AP MSC 62G0562N0162P10
keywords Huntington diseaseoutcome-dependent censoringright-censored covariatedoubly robust estimationsemiparametric efficiencynonparametric nuisance estimationclinical trial sample sizes
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

Clinical trials for Huntington disease need endpoints that change fast as participants approach Stage 1, but many participants drop out early, and estimators that ignore the dropout-outcome link give contradictory, sometimes biologically absurd slope estimates. SPYCE is an estimator for this right-censored-covariate problem that stays consistent when either the model for time to Stage 1 or the model for time to study exit is correctly specified, making it the first doubly robust estimator for outcome-dependent censoring of this kind. When both models are correct, SPYCE reaches the smallest possible asymptotic variance, and the same efficiency holds when both nuisance models are estimated nonparametrically, which the paper achieves by reformulating those models so they no longer depend on the unknown slope. Applied to PREDICT-HD, SPYCE resolves the contradiction, identifies caudate and putamen volume ratios as the most sensitive endpoints, and lowers required sample sizes to hundreds per arm. The cost is a single conditional-independence assumption: among participants with equal measured outcome and baseline covariates, exit time carries no extra information about time to Stage 1.

Core claim

Under outcome-dependent censoring, the paper proposes SPYCE to estimate the slope parameters β in the outcome model f_{Y|X,Z} via an estimating equation built from the efficient score function. The central claim is that if either the time-to-Stage-1 model f_{X|Z} or the censoring model f_{C|Y,Z} is correctly specified, SPYCE is consistent; if both are correct, it attains the semiparametric efficiency bound. A reformulation of the nuisance expectation operators removes their dependence on β, so both can be estimated nonparametrically, and the paper claims the same efficiency bound is still attained (Theorem 3(iii)). The paper asserts that no existing estimator achieved both double robustness

What carries the argument

The central object is the efficient score function S_eff, constructed as the projection of the outcome-model score onto the orthogonal complement of the nuisance tangent space; this estimating equation is exactly unbiased when either nuisance model is correctly specified, which is what delivers double robustness. The second load-bearing piece is the reformulation of the two conditional expectation operators E1 and E2 so they condition only on nuisance models and observed data, breaking the circular dependence on β that otherwise blocks nonparametric estimation. The Hájek-type normalization inside E2 is what lets the fully nonparametric Case 3 reach the semiparametric efficiency bound automat

Load-bearing premise

The load-bearing premise is that, given the measured outcome and baseline covariates, a participant's time to study exit carries no additional information about when they reach Stage 1; if early exit is driven by signs of approaching Stage 1 that the measured outcome does not capture, the estimator has no consistency guarantee.

What would settle it

Simulate a cohort as in Section 4 but add an unmeasured frailty U that accelerates time to Stage 1 and also shortens time to study exit, while keeping the distribution of the observed Y,Z identical; if SPYCE's slope estimates become biased or the 95% confidence-interval coverage for β2 drops below nominal, the identifying conditional independence C ⊥ X | Y,Z is the failure point.

Watch this falsifier — get emailed when new claim-graph text bears on it.

If this is right

  • SPYCE gives consistent slope estimates when either the time-to-Stage-1 model or the censoring model is correctly specified, so researchers no longer have to stake the analysis on which of two difficult models is right.
  • When both nuisance models are correct, the asymptotic variance reaches the semiparametric efficiency bound, which translates directly to the smallest sample sizes needed for a target precision.
  • Both nuisance models can be estimated nonparametrically without sacrificing efficiency, so the estimator remains reliable even when neither model can be parametrically specified.
  • Standard errors that account for nuisance-model estimation uncertainty are valid, preventing trials from being designed around overconfident slope estimates.
  • Applied to PREDICT-HD, the estimator reverses the ranking of candidate endpoints and brings required sample sizes per arm from hundreds of thousands to 241 for caudate volume ratio.

Where Pith is reading between the lines

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

  • Editorial inference: the conditional-independence structure assumed here appears in any longitudinal study where dropout tracks progression, such as Parkinson disease or spinocerebellar ataxia, so the estimator's machinery should transfer; a natural next step is external-validation replication on those cohorts.
  • Editorial inference: the reformulation that breaks the β-circularity is a reusable template for other semiparametric settings with censored covariates and outcome-dependent censoring, beyond the specific outcome model used in the paper.
  • Editorial inference: if the PREDICT-HD rankings hold, past endpoint-selection analyses that used complete-case or imputation estimators may have systematically deprioritized neuroimaging outcomes; reanalysis of published early-HD trial data could test this directly.
  • Editorial inference: a practical extension would be a data-adaptive bandwidth selector or cross-fitting version of the nonparametric SPYCE, which could reduce the modest undercoverage observed for Case 3 at high censoring rates.

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

3 major / 4 minor

Summary. The paper proposes SPYCE, a semiparametric estimator for the slope parameters β in the outcome model f_{Y|X,Z}(y,x,z;β), where the covariate X (time to Stage 1) is right-censored by a study-exit time C that may depend on the outcome Y. The model is the likelihood (1), which encodes the conditional independence C⊥X|Y,Z. SPYCE is shown to be doubly robust: consistent if either the time-to-event model f_{X|Z} or the censoring model f_{C|Y,Z} is correctly specified. With parametric nuisance models, the estimator is asymptotically normal and efficiency is achieved when both models are correct (Theorem 2). With nonparametric nuisance estimation, three cases are distinguished; the paper's headline claim is Theorem 3(iii), that when both expectation operators are estimated nonparametrically, the asymptotic variance automatically equals the semiparametric efficiency bound. The method is applied to PREDICT-HD data, resolving contradictory estimates and producing sample-size calculations for early-HD trials.

Significance. If the theoretical results are correct, SPYCE is a substantial contribution: it would be the first doubly robust, semiparametrically efficient estimator for this right-censored-covariate problem with outcome-dependent censoring, and the nonparametric Case 3 efficiency result is surprising and practically important. The paper includes detailed proofs in Section B, simulations matching PREDICT-HD censoring rates, and an honest discussion of the both-misspecified case where bias remains. The empirical application to identify sensitive endpoints is compelling. However, the central claims rest on the untestable assumption C⊥X|Y,Z, and the proof of the most novel theoretical result, Theorem 3(iii), has gaps that need attention. The practical sample-size claims depend on Case 3 using n=448 with 58% censoring, where the simulation coverage is already below nominal.

major comments (3)
  1. [§2.1, Eq. (1); Abstract] All consistency and efficiency guarantees are proved only under the conditional independence C⊥X|Y,Z, which is the factorization assumption behind likelihood (1). Dropout in neurodegenerative trials is plausibly driven by unmeasured progression that affects both Y and time to Stage 1 beyond what Y and Z capture. In that case (1) is not the true model, the efficient score in Proposition 2(iii) is not the efficient score for the true model, and Theorem 1's key step breaks. The abstract's claim that 'No existing estimator achieves both properties under outcome-dependent censoring' therefore overstates the scope: the paper addresses outcome-dependent censoring of the specific form C⊥X|Y,Z, not arbitrary dropout related to outcome. Please qualify the headline claim and, since the PREDICT-HD analysis has 58.3% censoring, provide a sensitivity analysis or at least a careful statement of the unt
  2. [§3.2.3, Theorem 3(iii), Proposition 3, and §B.6.4] The automatic efficiency preservation in Case 3 is the most surprising and load-bearing theoretical claim. In the proof, the nonparametric perturbation (pE1,pE2) is split into a Case-1 term with E2=E20 and a Case-2 term with E1=E10, and Proposition 3 is invoked to cancel the first-order terms. Because the equation defining a in (7) is nonlinear in (E1,E2), the cross term between (pE1-E10) and (pE2-E20) must be shown to be o_p(n^{-1/2}) uniformly; this is asserted via Lemmas B.3 and B.4 but not explicitly demonstrated. Moreover, Lemma B.4 claims invertibility of the linear operator L from injectivity alone; surjectivity and bounded inverse are not established. As written, the proof of Theorem 3(iii) is incomplete, and the central 'nonparametric without sacrificing efficiency' result requires a tighter argument or additional regularity conditions.
  3. [§4, Table 2; §5, Table 3] The PREDICT-HD sample-size claims, notably the 241 participants per arm for caudate volume ratio, are based on SPYCE-Non standard errors. In the simulation most similar to the application (high censoring, n=1,000), Case 3 has coverage 91.2% for β1, and the effective PREDICT-HD sample is only 448 with 58.3% censoring. The paper reports 'modest undercoverage', but 91.2% at n=1,000 suggests the practical standard errors may be optimistic at n≈448. Please provide evidence on finite-sample variance estimation at the PREDICT-HD sample size or temper the sample-size claims accordingly.
minor comments (4)
  1. [§2.1] The text 'meaning C |ù X|Y, Z' is typeset in an unusual way; the intended conditional independence C ⊥ X | (Y,Z) should be written explicitly, as this is the defining assumption of the model.
  2. [Abstract and §1] The claim that 'No existing estimator achieves both properties under outcome-dependent censoring' should be qualified as 'under the model (1) with C⊥X|Y,Z', to avoid implying a result for all forms of outcome-dependent censoring.
  3. [§3.2.2, Eq. (12)] The notation pE2p1|y,zq is ambiguous; since pE2 is an expectation operator, it would be clearer to write pE2{1|y,z} or pE2[1](y,z), and similarly elsewhere.
  4. [Table 2] In the 'Mis Mis' row under high censoring, the near-nominal coverage is correctly explained as an artifact of wide intervals. It would help to report the average interval length as well as coverage, so readers can distinguish robustness from scarcity-driven variability.

Circularity Check

0 steps flagged

No significant circularity: SPYCE's double robustness and efficiency claims are derived from a self-contained semiparametric calculation, with the paper explicitly identifying and resolving the one genuinely circular dependence in the nonparametric construction.

full rationale

The paper's derivation chain is self-contained. The target parameter β is identified from the observed-data likelihood (1) under the stated conditional independence C ⊥ X | Y, Z; Proposition 1 proves identifiability from the observed data, not from the estimator. Proposition 2 derives the nuisance tangent spaces and the efficient score Seff by direct orthogonality calculations within that semiparametric model. Theorem 1's double robustness is proven by showing E[S_eff] = 0 when either η1 or η2 is correctly specified, which is a standard doubly-robust estimating-equation argument and does not assume the conclusion. Theorems 2 and 3 account for nuisance-parameter estimation uncertainty through influence functions and variability functions, again by explicit asymptotic expansion. The only apparent circular dependence in the paper is in Section 3.2.2, where the authors state that estimating E1 and E2 nonparametrically requires knowing β, but estimating β requires E1 and E2; they then resolve this by reformulating the operators in (8)–(9) so that pE1 and pE2 can be evaluated at any candidate β without re-estimation. The Case 3 variance simplification to the semiparametric efficiency bound is derived, not assumed, via the cancellation established in Proposition 3. The self-citations (Lee et al. 2026; Zhang et al. 2025) are used only as prior-art contrasts requiring outcome-independent censoring and are not load-bearing for SPYCE's properties. The assumption C ⊥ X | Y, Z is an untestable identifying assumption that limits robustness to other forms of outcome-dependent dropout, but that is a model assumption, not a circular reduction of the derivation to its own output. Accordingly, no circular step is identified.

Axiom & Free-Parameter Ledger

4 free parameters · 6 axioms · 0 invented entities

The core estimator is derived from semiparametric first principles rather than from a fitted entity. The main auxiliary inputs are the two nuisance models (f_{X|Z} and f_{C|Y,Z}), the overlap support condition, and a battery of regularity conditions. The nonparametric case also selects bandwidths and kernel orders by hand, which are not reported. No new physical or probabilistic entity is postulated beyond the estimator itself.

free parameters (4)
  • Bandwidths h1, h2, h3 = not reported
    Bandwidths for the conditional Kaplan-Meier estimator and the kernel-based expectation operators are chosen by hand; the theory only constrains their rates via Conditions (N2)/(N21). No data-driven selection rule is given in the paper.
  • Kernel order m = not reported
    A higher-order kernel of order m > a/2 is required; the specific m used in the PREDICT-HD analysis or simulations is not reported, and results may depend on it.
  • Nuisance model parameters in SPYCE-Par (α11, α12, τ1², α21, α22, α23, τ2²) = MLEs from PREDICT-HD (values not shown)
    For the parametric version of SPYCE, the fitted parameters of the truncated normal models for X|Z and C|Y,Z affect the slope estimates and sample-size calculations. These are fitted to the same data, not independently derived.
  • Survival probability lower bound n^{-1} = n^{-1}
    The conditional Kaplan-Meier estimates are clipped at n^{-1} to prevent numerical instability; this is a hand-chosen constant that can affect small-sample behavior.
axioms (6)
  • domain assumption Conditional independence: time to study exit C is independent of time to Stage 1 X given outcome Y and baseline Z (C ⊥ X | Y,Z).
    This is the outcome-dependent censoring structure that makes likelihood (1) valid. If exit also depends directly on time to Stage 1 beyond Y, the likelihood and the double-robustness theorem do not apply. Stated in Section 2.1.
  • domain assumption Overlapping support condition: S_X|Y,Z(t)>0 iff S_C|Y,Z(t)>0 for all (t,y,z).
    Proposition 1 requires this for identifiability; without it, ranges of X beyond the maximum censoring time are unidentifiable.
  • domain assumption Regularity conditions for consistency and asymptotic normality: (C1)-(C4), (P1)-(P6), and for the nonparametric case (N1), (N2)/(N21), (N3)-(N7).
    These supremum boundedness, differentiability, and bandwidth conditions are standard but are load-bearing for the n^{1/2}-consistency and efficiency claims, especially the strong rate condition (N21).
  • domain assumption Correct specification of the outcome model f_{Y|X,Z} containing β.
    SPYCE estimates the slope β in the outcome model; misspecification of f_{Y|X,Z} would bias β even if the nuisance models are correct. In the data application this is a Normal linear model with interaction.
  • standard math Semiparametric theory background: nuisance tangent spaces, orthogonal complements, efficient influence functions.
    The derivation of Λ, Λ^⊥, and S_eff relies on the framework of Bickel et al. (1993) and Tsiatis (2006) as background.
  • domain assumption PREDICT-HD data and HD-ISS staging assumptions.
    The applied conclusions assume the PREDICT-HD cohort is representative, the imaging-based Stage 1 determination is valid, and the CAP>368 high-risk cut-off is appropriate. These are taken from the HD literature.

pith-pipeline@v1.3.0-alltime-deepseek · 75289 in / 12430 out tokens · 140521 ms · 2026-08-01T15:12:00.968272+00:00 · methodology

0 comments
read the original abstract

Clinical trials for neurodegenerative diseases must identify sensitive endpoints -- outcomes that change rapidly enough to detect treatment effects. In Huntington disease, this requires measuring how outcomes change as participants approach Stage 1. Yet many participants exit studies before reaching this stage, making their time to Stage 1 right-censored. Estimating how outcomes change requires models for both time to Stage 1 and time to study exit. When participants with worse outcomes exit earlier, this outcome-dependent censoring causes existing estimators to produce contradictory results: for the same cognitive outcome, one estimator suggests improvement while another shows decline. Existing estimators either ignore outcome-dependent censoring or require one model to be correctly specified, with no protection when it is not. We introduce SPYCE, a doubly robust estimator (consistent when either model is correctly specified) that achieves the smallest possible variance and allows both models to be estimated nonparametrically without sacrificing efficiency. Applied to data from PREDICT-HD, an observational Huntington disease study, SPYCE resolves current contradictions, identifies caudate and putamen volume ratios as the most promising sensitive endpoints, and shows that as few as 241 participants per arm are needed to detect treatment effects, versus hundreds of thousands under estimators that cannot handle outcome-dependent censoring.

Figures

Figures reproduced from arXiv: 2607.18501 by Karen Marder, Kihyun Han, Tanya P. Garcia, Yanyuan Ma.

Figure 1
Figure 1. Figure 1: Contradictory estimates of how rapidly Stroop Color Word Test scores change as [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Slope estimates and 95% confidence intervals for six outcomes in the high-risk [PITH_FULL_IMAGE:figures/full_fig_p019_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Power curves for detecting 50% treatment slowing in the high-risk group, assuming [PITH_FULL_IMAGE:figures/full_fig_p020_3.png] view at source ↗

discussion (0)

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

Reference graph

Works this paper leans on

64 extracted references · 1 linked inside Pith

  1. [1]

    and Jimenez-Maggiora, G.A

    Aisen, P.S. and Jimenez-Maggiora, G.A. and Rafii, M.S. and others , title=. Nature Reviews Neurology , volume=

  2. [2]

    and Marini, K

    Mahlknecht, P. and Marini, K. and Werkmann, M. and et al. , title=. Translational Neurodegeneration , volume=

  3. [3]

    and others , title=

    Tabrizi, S.J. and others , title=. Lancet Neurology , year=

  4. [4]

    and Darweesh, S.K.L

    Smedinga, M. and Darweesh, S.K.L. and Bloem, B.R. and Post, B. and Richard, E. , title=. Journal of Neurology , year=

  5. [5]

    Nature , year =

    Dolgin, Elie , title =. Nature , year =

  6. [6]

    Handbook of Econometrics , volume=

    Large sample estimation and hypothesis testing , author=. Handbook of Econometrics , volume=. 1994 , publisher=

    1994

  7. [7]

    Statistics in Medicine , volume =

    Jiang, Hui and Huang, Lei and Xia, Yingcun , title =. Statistics in Medicine , volume =

  8. [8]

    Journal of the American Statistical Association , volume=

    Locally weighted censored quantile regression , author=. Journal of the American Statistical Association , volume=. 2009 , publisher=

    2009

  9. [9]

    2012 , publisher=

    Nonparametric regression analysis of longitudinal data , author=. 2012 , publisher=

    2012

  10. [10]

    Beran, Rudolf , title =

  11. [11]

    Scandinavian Journal of Statistics , pages=

    Non-parametric regression with censored survival time data , author=. Scandinavian Journal of Statistics , pages=. 1987 , publisher=

    1987

  12. [12]

    Annals of Statistics , volume=

    Average treatment effects in the presence of unknown interference , author=. Annals of Statistics , volume=. 2021 , publisher=

    2021

  13. [13]

    Biometrika , volume=

    On inverse probability-weighted estimators in the presence of interference , author=. Biometrika , volume=. 2016 , publisher=

    2016

  14. [14]

    Statistics in Medicine , volume=

    Regression with a right-censored predictor using inverse probability weighting methods , author=. Statistics in Medicine , volume=. 2020 , publisher=

    2020

  15. [15]

    Disease progression in

    Garcia, Tanya P and Wang, Yuanjia and Shoulson, Ira and Paulsen, Jane S and Marder, Karen , journal=. Disease progression in. 2018 , publisher=

    2018

  16. [16]

    Clinical markers of early disease in persons near onset of

    Paulsen, Jane S and Zhao, H and Stout, JC and Brinkman, RR and Guttman, M and Ross, CA and Como, P and Manning, C and Hayden, MR and Shoulson, I and others , journal=. Clinical markers of early disease in persons near onset of. 2001 , publisher=

    2001

  17. [17]

    Tracking motor impairments in the progression of

    Long, Jeffery D and Paulsen, Jane S and Marder and others , journal=. Tracking motor impairments in the progression of. 2014 , publisher=

    2014

  18. [18]

    Indexing disease progression at study entry with individuals at-risk for

    Zhang, Ying and Long, Jeffrey D and Mills, James A and Warner, John H and Lu, Wenjing and others , journal=. Indexing disease progression at study entry with individuals at-risk for. 2011 , publisher=

    2011

  19. [19]

    arXiv preprint arXiv:2409.04684 , year=

    Establishing the parallels and differences between right-censored and missing covariates , author=. arXiv preprint arXiv:2409.04684 , year=

    Pith/arXiv arXiv

  20. [20]

    and Richardson, B

    Lee, S. and Richardson, B. D. and Ma, Y. and Marder, K. S. and Garcia, T. P. , title =. Journal of the American Statistical Association , volume =. 2026 , publisher =

    2026

  21. [21]

    Survival models and health sequences

    Kalbfleisch, John , year =. Discussion of “Survival models and health sequences” by Walter Dempsey and Peter McCullagh , volume =

  22. [22]

    arXiv preprint arXiv:2511.02187 , year=

    Super doubly robust and efficient estimator for informative covariate censoring , author=. arXiv preprint arXiv:2511.02187 , year=

    arXiv

  23. [23]

    Statistics in Medicine , volume=

    Nonparametric regression with right-censored covariate via conditional density function , author=. Statistics in Medicine , volume=. 2022 , publisher=

    2022

  24. [24]

    Journal of the Korean Statistical Society , volume=

    Maximum weighted likelihood for discrete choice models with a dependently censored covariate , author=. Journal of the Korean Statistical Society , volume=. 2017 , publisher=

    2017

  25. [25]

    Journal of Applied Statistics , volume=

    Multiple imputation of a randomly censored covariate improves logistic regression analysis , author=. Journal of Applied Statistics , volume=. 2016 , publisher=

    2016

  26. [26]

    Statistical Methods in Medical Research , volume=

    Improved conditional imputation for linear regression with a randomly censored predictor , author=. Statistical Methods in Medical Research , volume=. 2019 , publisher=

    2019

  27. [27]

    Biostatistics and Biometrics Open Access Journal , author=

    Linear Regression Model with a Randomly Censored Predictor: Estimation Procedures , year=. Biostatistics and Biometrics Open Access Journal , author=

  28. [28]

    Making sense of censored covariates: statistical methods for studies of

    Lotspeich, Sarah C and Ashner, Marissa C and Vazquez, Jesus E and Richardson, Brian D and Grosser, Kyle F and Bodek, Benjamin E and Garcia, Tanya P , journal=. Making sense of censored covariates: statistical methods for studies of. 2024 , publisher=

    2024

  29. [29]

    Applying the

    Long, Jeffrey D and Gantman, Emily C and Mills, James A and Vaidya, Jatin G and Mansbach, Alexandra and Tabrizi, Sarah J and Sampaio, Cristina , journal=. Applying the. 2023 , publisher=

    2023

  30. [30]

    Lifetime Data Analysis , volume=

    Survival models and health sequences , author=. Lifetime Data Analysis , volume=. 2018 , publisher=

    2018

  31. [31]

    Neuropathological classification of

    Vonsattel, Jean-Paul and Myers, Richard H and Stevens, Thomas J and Ferrante, Robert J and Bird, Edward D and Richardson Jr, Edward P , journal=. Neuropathological classification of. 1985 , publisher=

    1985

  32. [32]

    Scarpina, Federica and Tagini, Sofia , journal=. The. 2017 , publisher=

    2017

  33. [33]

    Clustering and prediction of disease progression trajectories in

    Ko, Jinnie and Furby, Hannah and Ma, Xiaoye and Long, Jeffrey D and Lu, Xiao-Yu and Slowiejko, Diana and Gandhy, Rita , journal=. Clustering and prediction of disease progression trajectories in. 2023 , publisher=

    2023

  34. [34]

    Linear regression with a randomly censored covariate: application to an

    Atem, Folefac D and Qian, Jing and Maye, Jacqueline E and Johnson, Keith A and Betensky, Rebecca A , journal=. Linear regression with a randomly censored covariate: application to an. 2017 , publisher=

    2017

  35. [35]

    Tracking

    Abeyasinghe, Pubu M and Long, Jeffrey D and Razi, Adeel and Pustina, Dorian and Paulsen, Jane S and Tabrizi, Sarah J and Poudel, Govinda R and Georgiou-Karistianis, Nellie , journal=. Tracking. 2021 , publisher=

    2021

  36. [36]

    Clinical and biomarker changes in premanifest

    Paulsen, Jane S and Long, Jeffrey D and Johnson, Hans J and Aylward, Elizabeth H and Ross, Christopher A and Williams, Janet K and Nance, Martha A and Erwin, Cheryl J and Westervelt, Holly J and Harrington, Deborah L and others , journal=. Clinical and biomarker changes in premanifest. 2014 , publisher=

    2014

  37. [37]

    Potential endpoints for clinical trials in premanifest and early

    Tabrizi, Sarah J and Reilmann, Ralf and Roos, Raymund AC and Durr, Alexandra and Leavitt, Blair and Owen, Gail and Jones, Rebecca and Johnson, Hans and Craufurd, David and Hicks, Stephen L and others , journal=. Potential endpoints for clinical trials in premanifest and early. 2012 , publisher=

    2012

  38. [38]

    Time-restricted eating in early-stage

    Wells, Russell G and Neilson, Lee E and McHill, Andrew W and Hiller, Amie L , journal=. Time-restricted eating in early-stage. 2025 , publisher=

    2025

  39. [39]

    Scandinavian Journal of Statistics , volume=

    Stochastic functional estimates in longitudinal models with interval-censored anchoring events , author=. Scandinavian Journal of Statistics , volume=. 2020 , publisher=

    2020

  40. [40]

    Biometrics , volume=

    Distribution-free estimation of local growth rates around interval censored anchoring events , author=. Biometrics , volume=. 2019 , publisher=

    2019

  41. [41]

    The dynamics of cognitive decline toward

    Kang, Kaidi and Zhang, Panpan and Dumitrescu, Logan and Mukherjee, Shubhabrata and Lee, Michael L and others , journal=. The dynamics of cognitive decline toward. 2025 , publisher=

    2025

  42. [42]

    Detection of

    Paulsen, Jane S and Langbehn, Douglas R and Stout, Julie C and Aylward, Elizabeth and Ross, Christopher A and Nance, Martha and Guttman, Mark and Johnson, Shannon and MacDonald, M and Beglinger, Leigh J and others , journal=. Detection of. 2008 , publisher=

    2008

  43. [43]

    Mild cognitive impairment in prediagnosed

    Duff, Kevin and Paulsen, Jane and Mills, J and Beglinger, LJ and Moser, DJ and Smith, MM and Langbehn, Douglas and Stout, Julie and Queller, Sarah and Harrington, DL , journal=. Mild cognitive impairment in prediagnosed. 2010 , publisher=

    2010

  44. [44]

    Preparing for preventive clinical trials: the

    Paulsen, Jane S and Hayden, Michael and Stout, Julie C and Langbehn, Douglas R and Aylward, Elizabeth and Ross, Christopher A and Guttman, Mark and Nance, Martha and Kieburtz, Karl and Oakes, David and others , journal=. Preparing for preventive clinical trials: the. 2006 , publisher=

    2006

  45. [45]

    Prediction of manifest

    Paulsen, Jane S and Long, Jeffrey D and Ross, Christopher A and Harrington, Deborah L and others , journal=. Prediction of manifest. 2014 , publisher=

    2014

  46. [46]

    Biological and clinical changes in premanifest and early stage

    Tabrizi, Sarah J and Scahill, Rachael I and Durr, Alexandra and Roos, Raymund AC and Leavitt, Blair R and Jones, Rebecca and Landwehrmeyer, G Bernhard and Fox, Nick C and Johnson, Hans and Hicks, Stephen L and others , journal=. Biological and clinical changes in premanifest and early stage. 2011 , publisher=

    2011

  47. [47]

    Neurology , author =

    Motor, cognitive, and functional declines contribute to a single progressive factor in early. Neurology , author =. 2017 , pages =

    2017

  48. [48]

    Earliest functional declines in

    Beglinger, Leigh J and O'Rourke, Justin JF and Wang, Chiachi and Langbehn, Douglas R and Duff, Kevin and Paulsen, Jane S and Huntington Study Group Investigators and others , journal=. Earliest functional declines in. 2010 , publisher=

    2010

  49. [49]

    Biometrics , volume=

    Threshold regression to accommodate a censored covariate , author=. Biometrics , volume=. 2018 , publisher=

    2018

  50. [50]

    Environmental and Ecological Statistics , volume=

    Handling covariates subject to limits of detection in regression , author=. Environmental and Ecological Statistics , volume=. 2012 , publisher=

    2012

  51. [51]

    Frontiers in Oncology , volume=

    Cancer survival: left truncation and comparison of results from hospital-based cancer registry and population-based cancer registry , author=. Frontiers in Oncology , volume=

  52. [52]

    Biometrika , volume=

    Mean squared errors of estimates of a density and its derivatives , author=. Biometrika , volume=. 1979 , publisher=

    1979

  53. [53]

    Scandinavian Journal of Statistics , pages=

    Estimating regression functions and their derivatives by the kernel method , author=. Scandinavian Journal of Statistics , pages=. 1984 , publisher=

    1984

  54. [54]

    Journal of the Royal Statistical Society

    Kernels for nonparametric curve estimation , author=. Journal of the Royal Statistical Society. Series B (Methodological) , pages=. 1985 , publisher=

    1985

  55. [55]

    1993 , publisher=

    Efficient and adaptive estimation for semiparametric models , author=. 1993 , publisher=

    1993

  56. [56]

    2006 , publisher=

    Semiparametric theory and missing data , author=. 2006 , publisher=

    2006

  57. [57]

    Movement Disorders , volume=

    Clinical outcomes and selection criteria for prodromal Huntington's disease trials , author=. Movement Disorders , volume=. 2020 , publisher=

    2020

  58. [58]

    Tominersen in adults with manifest

    McColgan, Peter and Thobhani, Alpa and Boak, Lauren and others , journal=. Tominersen in adults with manifest. 2023 , publisher=

    2023

  59. [59]

    Sathe, Swati and Ware, Jen and Levey, Jamie and Neacy, Eileen and Blumenstein, Robi and Noble, Simon and M. Enroll-. Frontiers in Neurology , volume=. 2021 , publisher=

    2021

  60. [60]

    A novel gene containing a trinucleotide repeat that is expanded and unstable on

    MacDonald, Marcy E and Ambrose, Christine M and Duyao, Mabel P and Myers, Richard H and Lin, Carol and Srinidhi, Lakshmi and Barnes, Glenn and Taylor, Sherryl A and James, Marianne and Groot, Nicolet and others , journal=. A novel gene containing a trinucleotide repeat that is expanded and unstable on. 1993 , publisher=

    1993

  61. [61]

    A biological classification of

    Tabrizi, Sarah J and Schobel, Scott and Gantman, Emily C and Mansbach, Alexandra and Borowsky, Beth and Konstantinova, Pavlina and Mestre, Tiago A and Panagoulias, Jennifer and Ross, Christopher A and Zauderer, Maurice and others , journal=. A biological classification of. 2022 , publisher=

    2022

  62. [62]

    An essay on the logical foundations of survey sampling, part one

    Comment on “An essay on the logical foundations of survey sampling, part one” , author=. Foundations of Statistical Inference , volume=. 1971 , publisher=

    1971

  63. [63]

    Gonzalez-Manteiga and C

    W. Gonzalez-Manteiga and C. Cadarso-Suarez , title =. Journal of Nonparametric Statistics , volume =. 1994 , publisher =

    1994

  64. [64]

    Uniform consistency of the kernel conditional

    Dabrowska, Dorota M , journal =. Uniform consistency of the kernel conditional