Pith. sign in

REVIEW 2 major objections 10 minor 298 references

A single extended-rank method handles regression and prediction for any ordinal outcome without specifying the transformation or deciding continuous versus discrete.

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 · grok-4.5

2026-07-31 03:45 UTC pith:ODUKLGXM

load-bearing objection Solid unification of rank regression for mixed ordinal data, with a real new binary-efficiency theorem and usable prediction tools; the conditional-coverage theory is one step removed from the intervals people will actually run. the 2 major comments →

arxiv 2607.25006 v1 pith:ODUKLGXM submitted 2026-07-27 stat.ME

Extended rank regression for all ordinal data

Peter Hoff , Supratik Basu This is my paper
classification stat.ME MSC 62J0562G0862F15
keywords rank likelihoodordinal datasemiparametric regressionconformal predictionconditional coverageBox–Cox transformationposterior predictionvariance-stabilizing transformation
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.

Regression accuracy hinges on the mean–variance relationship, yet that relationship is rarely the scientific target. This paper treats it as a pure nuisance by placing a nondecreasing transformation G on a normal linear predictor and basing all inference on an extended rank likelihood that depends only on the ordering (including ties) of the outcomes. The same procedure therefore applies to continuous, discrete, or mixed ordinal data and never requires estimating G or choosing a scale. At the two extremes—strictly continuous and binary data—the method is shown to lose no asymptotic efficiency relative to a correctly specified full likelihood, and rank-based prediction intervals achieve asymptotic coverage conditional on the features. Bayesian estimates and intervals come from a simple Gibbs sampler; when the model is doubted, conformal calibration of the rank predictive distribution still guarantees marginal frequentist coverage under exchangeability alone.

Core claim

The extended rank likelihood for a monotonically transformed linear model incurs no asymptotic information loss when estimating the regression coefficients at both extremes of the ordinal spectrum (continuous and binary), and the resulting rank-based prediction intervals attain asymptotic conditional coverage under the model, all without estimating or specifying the unknown transformation.

What carries the argument

The extended rank likelihood L(β : S(y)) = Pr(Z ∈ S(y) | β), where S(y) is the convex set of latent normal vectors consistent with the observed ordering and ties. It is free of G, is sampled by a simple Gibbs algorithm, and underpins both the efficiency theorems and the construction of rank-based (and conformal) prediction intervals.

Load-bearing premise

Outcomes must be a nondecreasing transformation of a normal linear predictor; if the latent errors are badly non-normal or the relationship is not monotone, the conditional-coverage claims no longer hold.

What would settle it

Generate continuous or binary data from a non-monotone or heavily non-normal latent process and check whether PERLE intervals still achieve the claimed conditional coverage and whether the estimator matches the efficiency of the full-likelihood MLE; or, on multi-category ordinal data, test whether efficiency equals that of a correctly specified ordered-probit MLE (only conjectured in the paper).

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

If this is right

  • One procedure can be applied to rainfall amounts, income brackets, Likert items, or continuous scores without deciding continuity or estimating a transformation.
  • Prediction-interval widths automatically adapt to the unknown mean–variance relationship induced by G.
  • When the monotone latent normal model is trusted, approximate conditional coverage is available; when it is not, the conformal rank procedure still guarantees marginal coverage under exchangeability.
  • The same efficiency argument is expected to extend to any finite number of ordered categories.

Where Pith is reading between the lines

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

  • The construction effectively converts any ordinal regression into a constrained multivariate-normal problem, so similar rank likelihoods could replace explicit transformation families in other generalized linear settings.
  • Because the conformal score is the posterior predictive probability of the latent rank, the same device can be attached to other Bayesian ordinal models to obtain ordering-respecting, model-robust intervals.
  • The Gibbs sampler’s simplicity may make the method competitive with standard ordered-probit software precisely when the number of categories is large or unknown in advance.

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 / 10 minor

Summary. The paper develops inference and prediction for the monotonically transformed linear model (MTLM), Y_i = G(Z_i) with Z ~ N(Xβ, I) and G an arbitrary nondecreasing nuisance function, using Pettitt's extended rank likelihood (ERL), which depends on the data only through min/max-rank intervals and hence accommodates continuous, binary, multi-category, and mixed ordinal outcomes with one method. A Gibbs sampler (PERLE) gives posterior inference for β. Theoretical contributions: an ERL score/information decomposition with a Brascamp–Lieb bound (Eq. 10, Lemma 2); a small result (Theorem 1) that at β=0 the ERL information matches the efficient information up to o_p(n); restatement of Bickel–Ritov's semiparametric efficiency of the MRLE (Theorem 2); a new and carefully proved result (Theorem 3) that for binary outcomes the ERL is an integrated probit likelihood and PERLE is asymptotically equivalent to the full probit MLE; and conditional-coverage theory (Lemma 3, Corollary 1, Theorems 4–5) for plug-in rank-based prediction intervals. Practical prediction intervals are built from a posterior predictive extended-rank distribution (§4.2) and from a conformal calibration thereof with exchangeability-based marginal coverage (§4.3). Two data analyses (Seattle rainfall, GSS income) show markedly better conditional coverage than transformed-linear, weighted-conformal, quantile-regression-conformal, and full ordinal probit baselines. Proofs are detailed and appear correct; code and an R (per

Significance. If the results hold, the paper delivers a single, transformation-free inferential and predictive framework for the full spectrum of ordinal data, backed by genuine asymptotic content rather than heuristics: semiparametric efficiency at the continuous extreme (Bickel–Ritov, plus the authors' Theorem 1 at β=0), a new efficiency proof at the binary extreme, Lipschitz control of the plug-in rank CDF (Lemma 3), and distribution-free marginal coverage from the conformal variant. The work ships with replication code and an R package, and the empirical claims (conditional coverage across outcome quantile bins in Figures 1, 4, 6) are directly falsifiable and well supported. The treatment of G as a nuisance parameter eliminates a real and common source of analyst arbitrariness (choice of transformation, discrete-vs-continuous decision). The limitations are clearly acknowledged: intermediate-K efficiency is conjectured, and conditional-coverage theory assumes the MTLM.

major comments (2)
  1. [§4.1–4.2, Theorems 4–5 vs. §4.2] The abstract's claim that "rank-based prediction intervals can obtain approximate coverage control conditional on the features" is proved (Theorems 4–5) only for a plug-in order-statistic procedure: choose γ̂_l, γ̂_u rank-measurable with H(γ̂_u, β̂) − H(γ̂_l, β̂) → 1−α for a √n-consistent β̂, then use (Z_(⌊γ̂_ln⌋), Z_(⌈γ̂_un⌉)). The interval actually proposed and used in §5 is the PERLE-Bayes interval of §4.2, whose l, u come from the posterior predictive rank distribution Pr(r(Z^{n+1})_{n+1} = k | Z∈S(y)). For this interval the paper proves only a Bayesian marginal statement (§4.2, final paragraph, event-containment argument) and itself notes it "is not a posterior prediction interval in the usual sense." No frequentist conditional-coverage result Pr(Y∈C_PERLE | x) → 1−α is established. The gap is presumably second-order (posterior concentration should make the integrated rank probabili
  2. [§4.1, Corollary 1 and Theorem 5] Corollary 1 (hence Theorems 4–5) assumes max{||x_1||,…,||x_n||} = O(1), while Theorem 5 simultaneously posits an i.i.d. design x_i ~ P_x. With only E||x_i||² < ∞ the max grows like o(n^{1/2}), not O(1); bounded support or a higher moment condition (E||x||^q < ∞, q>2, giving max = o_p(n^{1/q})) is needed for the plug-in error ||X(β−β̂)||∞ + |xᵀ(β−β̂)| to vanish at a fixed x. The remark after Theorem 5 acknowledges a relaxation but does not state the moment condition on P_x that makes it hold. Separately, Theorem 5's conclusion is pointwise in x; since Lemma 3's bound contains |xᵀ(β−β̂)|, the convergence is uniform only over x in compact sets, which matters for interpreting "conditional coverage for all x ∈ R^p" in §4.1. Please state the design/moment assumptions precisely and clarify the mode of uniformity in x.
minor comments (10)
  1. [§2.1] Typo: "depends on the data only though the ranks" should be "through".
  2. [§2.1, Eq. (4)] Eq. (4): the max and min in the definition of S(y) are over empty index sets when y_i is the sample minimum or maximum; state the ±∞ conventions explicitly.
  3. [§3, Theorem 3 preamble] In the binary-case ERL representation preceding Theorem 3, the pseudo-prior w(θ) depends on the data through N_0, so the "integrated likelihood" framing is nonstandard. The proof of Lemma 5 handles this, but a remark flagging it for readers would help.
  4. [§3, Theorem 1] The restriction to β=0 is motivated as a special case complementing Bickel–Ritov, but one sentence explaining what blocks the argument for general β (dependence between X and the ranks/order statistics) would clarify its role.
  5. [§4.2] The bullets defining y and ȳ via k_l = min{k : l ≤ s_k} and the extended quantile formula ˆF^{−1}(l/n) deserve a small worked check or remark: at l=0 or u=n+1 the interval endpoints fall back on y_(0), y_(K+1), "the smallest and largest possible y-values" — in practice these require a known support bound, which should be discussed.
  6. [§5.1, Figure 4] Figure 4 uses point glyphs (digits 1–5) for five methods across eight bins; it is hard to read. Consider connected lines or faceting, as in Figure 6.
  7. [§5.1] The conformal guarantee in §4.3 rests on exchangeability, but the rainfall data are a daily time series and the paper itself reports residual autocorrelation (lag-1 ≈ 0.056) after transformation. The rolling-origin evaluation is sensible, but a sentence noting that the marginal-coverage guarantee does not strictly apply under temporal dependence — and why the small autocorrelation makes the violation mild — would be appropriate.
  8. [§5.1] The Gaussian kernel bandwidth for the locally weighted conformal baselines was "chosen by trial and error." For reproducibility (and fairness of the comparison) report the value and selection criterion.
  9. [Appendix, proof of Lemma 2] The remark that Lemma 2's proof extends to Z ~ N(µ, Σ) giving Var[Z|Z∈S] ⪯ Σ is stated only inside the proof; it is a useful standalone corollary.
  10. [§4.3] The conformal procedure recomputes extended ranks and scores for each of 2K+1 candidate ranks; with K=1097 in the rainfall data this is potentially expensive. A comment on computational cost and any shortcuts used in perle would be valuable.

Circularity Check

0 steps flagged

No circularity: efficiency and coverage claims are derived from the ERL definition and standard asymptotic arguments, not from self-fitting or load-bearing self-citation.

full rationale

The paper’s load-bearing results are (i) no asymptotic information loss of the extended rank likelihood at the continuous extreme (Theorem 2, citing Bickel–Ritov 1997) and binary extreme (Theorem 3, proved in the appendix via integrated-probit representation and posterior–MLE equivalence), and (ii) asymptotic conditional coverage of a plug-in rank order-statistic interval under the MTLM (Theorems 4–5, via Lipschitz coupling of the rank CDF H and rank-measurability). These are ordinary likelihood/asymptotic derivations from the stated model Z~N(Xβ,I), Y=G(Z) and the definition L(β:S(y))=Pr(Z∈S(y)|β). Self-citations (Hoff 2007/2008 on the ERL and Gibbs sampler; Hoff 2023 on posterior-predictive conformity scores) supply computational tools and score motivation; none is invoked as a uniqueness theorem that forces the efficiency or coverage claims. Conformal marginal coverage follows from exchangeability of the rank scores, independent of MTLM truth. Prediction intervals use fitted β to forecast new Y—standard prediction, not “fitted input relabeled as prediction.” No step reduces a claimed derivation to its own inputs by construction. Score 0 is appropriate.

Axiom & Free-Parameter Ledger

3 free parameters · 6 axioms · 2 invented entities

Central results rest on the MTLM (latent i.i.d. normal linear scores, unknown nondecreasing G, fixed scale and no intercept), standard regularity for M-estimators/probit information, and exchangeability for conformal guarantees. Computational defaults (Gaussian prior scale, MCMC length) affect finite-sample output but not the asymptotic claims. No new physical entities; ‘extended ranks’ and ERL are definitions/statistics, not postulated mechanisms.

free parameters (3)
  • Prior scale τ² for β ~ N(0, τ² I) = Not numerically fixed in the text; standard ridge-style hyperparameter
    Enters the Gibbs full conditional and the PERLE posterior; asymptotic claims assume a nonsingular normal prior but finite-sample intervals depend on the chosen τ.
  • MCMC length / thinning (11k iterations, 1k burn-in, thin 10) = 11000 / 1000 / 10 as reported in §5
    Used for both empirical examples; affects Monte Carlo error of reported PERLEs and predictive rank probabilities.
  • Conformal split sizes and competitor kernel bandwidth = 365-day calibration; bandwidth trial-and-error
    Rain example uses preceding 365 days for calibration; weighted conformal bandwidth ‘chosen by trial and error’—affects baseline comparisons, not the PERLE theory.
axioms (6)
  • domain assumption MTLM: Z ~ N_n(Xβ, I_n), Y_i = G(Z_i) for unknown nondecreasing G; variance fixed at 1 and no intercept for identifiability
    Foundational model in §2.1; all likelihood, efficiency, and conditional-coverage results are under this specification.
  • domain assumption For continuous efficiency: regularity conditions of Bickel & Ritov (1997) for transformation models
    Invoked as Theorem 2; paper does not re-derive the full LAN argument.
  • domain assumption Binary case: i.i.d. (x_i, Y_i), E[||x_i||^6] < ∞, Var(x_i) positive definite; standard probit MLE asymptotics (Fahrmeir & Kaufmann)
    Assumptions stated in Theorem 3; used to transfer efficiency from full probit MLE to PERLE.
  • domain assumption Prediction asymptotics: √n-consistent β̂ and max_i ||x_i|| = O(1) (or o(√n)); optional i.i.d. design P_x for marginal conditional coverage
    Corollary 1 and Theorems 4–5 in §4.1.
  • standard math Conformal guarantee requires only exchangeability of {(Y_i, x_i)} including the test point
    Standard conformal argument applied to extended-rank scores in §4.3.
  • standard math Brascamp–Lieb variance inequality for log-concave densities (used to show Var[Z|Z∈S] ⪯ I)
    Lemma 2 proof approximates the truncated Gaussian by smooth strictly log-concave densities.
invented entities (2)
  • Extended ranks (min-rank / max-rank interval per observation) independent evidence
    purpose: Encode the partial order on latent Z implied by a non-strictly monotone G, including ties
    Definition 1; standardizes tied ordinal data so the ERL is Pr(r(Z)_i ∈ r(y)_i). Definitional statistic, not a latent physical object.
  • PERLE (posterior extended rank likelihood estimation) and PERLE-conformal intervals independent evidence
    purpose: Name the Gibbs posterior under the ERL and the conformal procedure that uses posterior rank probabilities as conformity scores
    Branding for the computational pipeline in §§2.2 and 4.2–4.3; substance is standard Bayesian missing-data MCMC plus split conformal.

pith-pipeline@v1.2.0-grok45-kimik3 · 33237 in / 3794 out tokens · 72464 ms · 2026-07-31T03:45:03.774753+00:00 · methodology

0 comments
read the original abstract

The accuracy of inference from a regression model depends largely on how well the model represents the relationship between the mean and variance of the outcomes. As this relationship is rarely of direct interest, it is natural to treat it as a nuisance parameter, rather than attempt to estimate it. We take this approach in the context of a monotonically transformed linear regression model using a pseudo-likelihood based on an extended notion of ranks. This approach can accommodate a wide range of mean-variance relationships and any ordinal data type, including continuous and discrete ordered data, and requires no estimation or prior specification of the transformation, or decision to treat an outcome as continuous or discrete. We show that the extended rank likelihood incurs no asymptotic information loss at the two extremes of continuous and binary data, and that rank-based prediction intervals can obtain approximate coverage control conditional on the features. Bayesian parameter estimates and prediction intervals are available via a simple Gibbs sampling algorithm. For settings where the model is in doubt, conformal calibration of the Bayesian predictive distribution provides intervals with guaranteed marginal frequentist coverage.

Figures

Figures reproduced from arXiv: 2607.25006 by Peter Hoff, Supratik Basu.

Figure 1
Figure 1. Figure 1: Seattle rain data. The left panel plots conditional coverage rates of nominal [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Seattle rainfall data from 2016-01-01 through 2017-12-31. The left panel [PITH_FULL_IMAGE:figures/full_fig_p023_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Posterior summary for the Seattle rain data. The left panel displays [PITH_FULL_IMAGE:figures/full_fig_p024_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Prediction intervals for the Seattle rainfall data. The left and right panels [PITH_FULL_IMAGE:figures/full_fig_p026_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Posterior summary for the income data. The left panel displays the PER [PITH_FULL_IMAGE:figures/full_fig_p027_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Prediction intervals for the income data. The left and right panels display [PITH_FULL_IMAGE:figures/full_fig_p028_6.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

298 extracted references · 77 canonical work pages · 1 internal anchor

  1. [1]

    Sampford, M. R. , title =. The Annals of Mathematical Statistics , volume =

  2. [2]

    Hothorn, Torsten and Kneib, Thomas and B\"uhlmann, Peter , TITLE =. J. R. Stat. Soc. Ser. B. Stat. Methodol. , FJOURNAL =. 2014 , NUMBER =. doi:10.1111/rssb.12017 , URL =

    work page doi:10.1111/rssb.12017 2014

  3. [3]

    Fahrmeir, Ludwig and Kaufmann, Heinz , TITLE =. Statist. Hefte (N.F.) , FJOURNAL =. 1986 , NUMBER =. doi:10.1007/BF02932567 , URL =

    work page doi:10.1007/bf02932567 1986

  4. [4]

    Bobkov, Sergey and Ledoux, Michel , TITLE =. Mem. Amer. Math. Soc. , FJOURNAL =. 2019 , NUMBER =. doi:10.1090/memo/1259 , URL =

    doi:10.1090/memo/1259 2019

  5. [5]

    , TITLE =

    H\'ajek, Jaroslav and Sid\'ak, Zbyn ek and Sen, Pranab K. , TITLE =. 1999 , PAGES =

    1999

  6. [6]

    , TITLE =

    Brascamp, Herm Jan and Lieb, Elliott H. , TITLE =. J. Functional Analysis , FJOURNAL =. 1976 , NUMBER =. doi:10.1016/0022-1236(76)90004-5 , URL =

    doi:10.1016/0022-1236(76)90004-5 1976

  7. [7]

    The validity of posterior expansions based on Laplace's method." Essays in Honor of George Barnard, eds. S. Geisser, JS Hodges , author=. 1990 , publisher=

    1990

  8. [8]

    2025 , note =

    probably: Tools for Post-Processing Predicted Values , author =. 2025 , note =

    2025

  9. [9]

    Conformalized Quantile Regression , url =

    Romano, Yaniv and Patterson, Evan and Candes, Emmanuel , booktitle =. Conformalized Quantile Regression , url =

  10. [10]

    Proceedings of the 13th European Conference on Machine Learning (ECML 2002) , editor =

    Papadopoulos, Harris and Proedrou, Kostas and Vovk, Volodya and Gammerman, Alexander , title =. Proceedings of the 13th European Conference on Machine Learning (ECML 2002) , editor =. 2002 , doi =

    2002

  11. [11]

    , title =

    Davern, Michael and Bautista, Rene and Freese, Jeremy and Herd, Pamela and Morgan, Stephen L. , title =. 2024 , publisher =

    2024

  12. [12]

    Biometrika , FJOURNAL =

    Guan, Leying , TITLE =. Biometrika , FJOURNAL =. 2023 , NUMBER =. doi:10.1093/biomet/asac040 , URL =

    doi:10.1093/biomet/asac040 2023

  13. [13]

    Weather and Forecasting , volume=

    Skill of global raw and postprocessed ensemble predictions of rainfall in the tropics , author=. Weather and Forecasting , volume=

  14. [14]

    Pettitt, A. N. , TITLE =. Biometrika , FJOURNAL =. 1984 , NUMBER =. doi:10.1093/biomet/71.1.35 , URL =

    work page doi:10.1093/biomet/71.1.35 1984

  15. [15]

    and Sabatti, Chiara , TITLE =

    Liu, Jun S. and Sabatti, Chiara , TITLE =. Biometrika , FJOURNAL =. 2000 , NUMBER =. doi:10.1093/biomet/87.2.353 , URL =

    work page doi:10.1093/biomet/87.2.353 2000

  16. [16]

    Bickel, P. J. , TITLE =. Proceedings of the 1st. 1987 , ISBN =. doi:10.1163/22134379-90003338 , URL =

    work page doi:10.1163/22134379-90003338 1987

  17. [17]

    , TITLE =

    Horowitz, Joel L. , TITLE =. Econometrica , FJOURNAL =. 1996 , NUMBER =. doi:10.2307/2171926 , URL =

    work page doi:10.2307/2171926 1996

  18. [18]

    Ye, Jianming and Duan, Naihua , TITLE =. Ann. Statist. , FJOURNAL =. 1997 , NUMBER =. doi:10.1214/aos/1030741091 , URL =

    arXiv 1997

  19. [19]

    Gu, Minggao and Wu, Yueqin and Huang, Bin , TITLE =. J. Multivariate Anal. , FJOURNAL =. 2014 , PAGES =. doi:10.1016/j.jmva.2013.08.016 , URL =

    work page doi:10.1016/j.jmva.2013.08.016 2014

  20. [20]

    Econometrica , FJOURNAL =

    Chen, Songnian , TITLE =. Econometrica , FJOURNAL =. 2002 , NUMBER =. doi:10.1111/1468-0262.00347 , URL =

    arXiv 2002

  21. [21]

    Signal Processing , volume =

    Pierre Comon , title =. Signal Processing , volume =. 1994 , issn =

    1994

  22. [22]

    , TITLE =

    Hoff, P.D. , TITLE =. J. Amer. Statist. Assoc. , FJOURNAL =. 2022 , NUMBER =. doi:10.1080/01621459.2020.1844720 , URL =. arXiv:arXiv:1907.12589 , code=

    arXiv 2022

  23. [23]

    arXiv preprint arXiv:2411.11824 , year=

    Theoretical foundations of conformal prediction , author=. arXiv preprint arXiv:2411.11824 , year=

    Pith/arXiv arXiv

  24. [24]

    Clarke, Bertrand and Rigo, Pietro , TITLE =. Statist. Sci. , FJOURNAL =. 2025 , NUMBER =. doi:10.1214/24-sts963 , URL =

    work page doi:10.1214/24-sts963 2025

  25. [25]

    , TITLE =

    Gleser, Leon Jay and Hwang, Jiunn T. , TITLE =. Ann. Statist. , FJOURNAL =. 1987 , NUMBER =. doi:10.1214/aos/1176350597 , URL =

    arXiv 1987

  26. [26]

    Gleser, Leon Jay , TITLE =. Ann. Statist. , FJOURNAL =. 1981 , NUMBER =

    1981

  27. [27]

    , TITLE =

    Fuller, Wayne A. , TITLE =. 1987 , PAGES =. doi:10.1002/9780470316665 , URL =

    doi:10.1002/9780470316665 1987

  28. [28]

    , TITLE =

    Keener, Robert W. , TITLE =. 2010 , PAGES =. doi:10.1007/978-0-387-93839-4 , URL =

    work page doi:10.1007/978-0-387-93839-4 2010

  29. [30]

    , TITLE =

    Tseng, Yu-Ling and Brown, Lawrence D. , TITLE =. Ann. Statist. , FJOURNAL =. 1997 , NUMBER =. doi:10.1214/aos/1069362396 , URL =

    arXiv 1997

  30. [31]

    and Searle, Shayle R

    McCulloch, Charles E. and Searle, Shayle R. and Neuhaus, John M. , TITLE =. 2008 , PAGES =

    2008

  31. [32]

    and Nelder, J

    McCullagh, P. and Nelder, J. A. , TITLE =. 1989 , PAGES =. doi:10.1007/978-1-4899-3242-6 , URL =

    doi:10.1007/978-1-4899-3242-6 1989

  32. [33]

    Linear Algebra Appl

    Burns, Fennell and Fiedler, Miroslav and Haynsworth, Emilie , TITLE =. Linear Algebra Appl. , FJOURNAL =. 1974 , NUMBER =. doi:10.1016/0024-3795(74)90089-5 , URL =

    work page doi:10.1016/0024-3795(74)90089-5 1974

  33. [34]

    Marshall, A. W. and Olkin, I. , TITLE =. Aequationes Math. , FJOURNAL =. 1990 , NUMBER =. doi:10.1007/BF02112284 , URL =

    work page doi:10.1007/bf02112284 1990

  34. [35]

    , title =

    Bryan, Jordan and Hoff, Peter and Osburn, Christopher L. , title =. ACS ES&T Water , volume =. 2023 , doi =

    2023

  35. [36]

    Technometrics , volume =

    Bryan, Jordan and Hoff, Peter and Osburn, Christopher , title =. Technometrics , volume =. 2025 , publisher =. doi:10.1080/00401706.2024.2379850 , URL =

    arXiv 2025

  36. [37]

    and Handsel, Lauren T

    Osburn, Christopher L. and Handsel, Lauren T. and Peierls, Benjamin L. and Paerl, Hans W. , title =. Environmental Science & Technology , volume =. 2016 , doi =

    2016

  37. [38]

    , TITLE =

    Smith, Richard L. , TITLE =. Biometrika , FJOURNAL =. 1994 , NUMBER =. doi:10.1093/biomet/81.1.173 , URL =

    work page doi:10.1093/biomet/81.1.173 1994

  38. [39]

    The Annals of Statistics , volume=

    Differential geometry of curved exponential families-curvatures and information loss , author=. The Annals of Statistics , volume=. 1982 , publisher=

    1982

  39. [40]

    2016 , PAGES =

    Amari, Shun-ichi , TITLE =. 2016 , PAGES =. doi:10.1007/978-4-431-55978-8 , URL =

    doi:10.1007/978-4-431-55978-8 2016

  40. [41]

    Sound analysis and synthesis with

    Sueur, J. Sound analysis and synthesis with. 2018 , publisher=

    2018

  41. [42]

    A Modern Gauss-Markov Theorem? Really?

    P\"otscher, Benedikt M. and Preinerstorfer, David , title =. 2022 , archivePrefix = "arXiv", primaryClass = "math.ST", eprint =. doi:10.48550/ARXIV.2203.01425 , url =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2203.01425 2022

  42. [43]

    Tan, W. Y. and Guttman, Irwin , TITLE =. J. Roy. Statist. Soc. Ser. B , FJOURNAL =. 1971 , PAGES =

    1971

  43. [44]

    , TITLE =

    Lee, John M. , TITLE =. 2018 , PAGES =

    2018

  44. [45]

    The American Statistician , volume =

    Stephen Portnoy , title =. The American Statistician , volume =. 2022 , publisher =

    2022

  45. [46]

    , TITLE =

    Higham, Nicholas J. , TITLE =. 2008 , PAGES =. doi:10.1137/1.9780898717778 , URL =

    doi:10.1137/1.9780898717778 2008

  46. [47]

    1999 , PAGES =

    Lang, Serge , TITLE =. 1999 , PAGES =. doi:10.1007/978-1-4612-0541-8 , URL =

    doi:10.1007/978-1-4612-0541-8 1999

  47. [48]

    2007 , PAGES =

    Bhatia, Rajendra , TITLE =. 2007 , PAGES =

    2007

  48. [49]

    Kaiser, Henry F , journal=. The. 1958 , publisher=

    1958

  49. [50]

    Yin, Jianxin and Li, Hongzhe , TITLE =. J. Multivariate Anal. , FJOURNAL =. 2012 , PAGES =. doi:10.1016/j.jmva.2012.01.005 , URL =

    work page doi:10.1016/j.jmva.2012.01.005 2012

  50. [51]

    Geodesic convexity in nonlinear optimization , JOURNAL =

    Rapcs\'. Geodesic convexity in nonlinear optimization , JOURNAL =. 1991 , NUMBER =. doi:10.1007/BF00940467 , URL =

    work page doi:10.1007/bf00940467 1991

  51. [52]

    Geodesic Convexity and Covariance Estimation , year=

    Wiesel, Ami , journal=. Geodesic Convexity and Covariance Estimation , year=

  52. [53]

    Anderson, T. W. , TITLE =. 2003 , PAGES =

    2003

  53. [54]

    Biometrika , FJOURNAL =

    Cacoullos, Theophilos and Olkin, Ingram , TITLE =. Biometrika , FJOURNAL =. 1965 , PAGES =. doi:10.2307/2333814 , URL =

    work page doi:10.2307/2333814 1965

  54. [55]

    Journal of the American Statistical Association , year =

    Nikolaos Ignatiadis and Stefan Wager , title =. Journal of the American Statistical Association , year =. doi:10.1080/01621459.2021.2008403 , URL =

    arXiv 2021

  55. [56]

    2021 , note =

    fabCI: FAB Confidence Intervals , author =. 2021 , note =

    2021

  56. [57]

    and Louis, Thomas A

    Laird, Nan M. and Louis, Thomas A. , TITLE =. J. Amer. Statist. Assoc. , FJOURNAL =. 1987 , NUMBER =

    1987

  57. [58]

    , TITLE =

    Morris, Carl N. , TITLE =. Scientific inference, data analysis, and robustness (. 1983 , MRCLASS =

    1983

  58. [59]

    Jiang, Wenhua and Zhang, Cun-Hui , TITLE =. Ann. Statist. , FJOURNAL =. 2009 , NUMBER =. doi:10.1214/08-AOS638 , URL =

    doi:10.1214/08-aos638 2009

  59. [60]

    Heinrich, Philippe and Kahn, Jonas , TITLE =. Ann. Statist. , FJOURNAL =. 2018 , NUMBER =. doi:10.1214/17-AOS1641 , URL =

    doi:10.1214/17-aos1641 2018

  60. [61]

    and Kalbfleisch, J

    Susko, E. and Kalbfleisch, J. D. and Chen, J. , TITLE =. Canad. J. Statist. , FJOURNAL =. 1998 , NUMBER =. doi:10.2307/3315720 , URL =

    work page doi:10.2307/3315720 1998

  61. [62]

    Kim, Arlene K. H. , TITLE =. Bernoulli , FJOURNAL =. 2014 , NUMBER =. doi:10.3150/13-BEJ542 , URL =

    work page doi:10.3150/13-bej542 2014

  62. [63]

    and Lesperance, Mary L

    Lindsay, Bruce G. and Lesperance, Mary L. , TITLE =. J. Statist. Plann. Inference , FJOURNAL =. 1995 , NUMBER =. doi:10.1016/0378-3758(94)00120-K , URL =

    work page doi:10.1016/0378-3758(94)00120-k 1995

  63. [64]

    1995 , PAGES =

    Lindsay, Bruce G , TITLE =. 1995 , PAGES =

    1995

  64. [65]

    , TITLE =

    Friedman, Jerome H. , TITLE =. J. Amer. Statist. Assoc. , FJOURNAL =. 1989 , NUMBER =

    1989

  65. [66]

    Green, P. J. , TITLE =. J. Roy. Statist. Soc. Ser. B , FJOURNAL =. 1984 , NUMBER =

    1984

  66. [67]

    , TITLE =

    Liang, Kung Yee and Zeger, Scott L. , TITLE =. Biometrika , FJOURNAL =. 1986 , NUMBER =. doi:10.1093/biomet/73.1.13 , URL =

    doi:10.1093/biomet/73.1.13 1986

  67. [68]

    Biometrics , FJOURNAL =

    Fiksel, Jacob and Zeger, Scott and Datta, Abhirup , TITLE =. Biometrics , FJOURNAL =. 2022 , NUMBER =. doi:10.1111/biom.13465 , URL =

    work page doi:10.1111/biom.13465 2022

  68. [69]

    , TITLE =

    Zeger, Scott L. , TITLE =. Biometrika , FJOURNAL =. 1988 , NUMBER =. doi:10.1093/biomet/75.4.621 , URL =

    work page doi:10.1093/biomet/75.4.621 1988

  69. [70]

    , TITLE =

    Greene, Tom and Rayens, William S. , TITLE =. Comm. Statist. Theory Methods , FJOURNAL =. 1989 , NUMBER =. doi:10.1080/03610928908830117 , URL =

    work page doi:10.1080/03610928908830117 1989

  70. [71]

    Rayens, William and Greene, Tom , TITLE =. Comput. Statist. Data Anal. , FJOURNAL =. 1991 , NUMBER =. doi:10.1016/0167-9473(91)90050-C , URL =

    work page doi:10.1016/0167-9473(91)90050-c 1991

  71. [72]

    Speech Recognition Using Articulatory and Excitation Source Features , author=

  72. [73]

    Uwe Ligges and Sebastian Krey and Olaf Mersmann and Sarah Schnackenberg , year =

  73. [74]

    Panaretos , title =

    Tomas Masak and Victor M. Panaretos , title =. Journal of the American Statistical Association , volume =. 2022 , publisher =. doi:10.1080/01621459.2022.2061982 , URL =

    arXiv 2022

  74. [75]

    and Sarkar, S

    Masak, T. and Sarkar, S. and Panaretos, V. M. , TITLE =. Biometrika , FJOURNAL =. 2023 , NUMBER =. doi:10.1093/biomet/asac035 , URL =

    work page doi:10.1093/biomet/asac035 2023

  75. [76]

    2017 , eprint=

    A representation theorem for stochastic processes with separable covariance functions, and its implications for emulation , author=. 2017 , eprint=

    2017

  76. [77]

    2018 , eprint=

    Speech Commands: A Dataset for Limited-Vocabulary Speech Recognition , author=. 2018 , eprint=

    2018

  77. [78]

    , author=

    Speech Commands: A public dataset for single-word speech recognition. , author=

  78. [79]

    , TITLE =

    Aerts, Marc and Claeskens, Gerda and Hart, Jeffrey D. , TITLE =. Ann. Statist. , FJOURNAL =. 2004 , NUMBER =. doi:10.1214/009053604000000805 , URL =

    work page doi:10.1214/009053604000000805 2004

  79. [80]

    , TITLE =

    Hart, Jeffrey D. , TITLE =. J. Stat. Theory Pract. , FJOURNAL =. 2009 , NUMBER =. doi:10.1080/15598608.2009.10411954 , URL =

    arXiv 2009

  80. [81]

    and Choi, Taeryon , TITLE =

    Hart, Jeffrey D. and Choi, Taeryon , TITLE =. Bayesian Anal. , FJOURNAL =. 2017 , NUMBER =. doi:10.1214/16-BA1018 , URL =

    work page doi:10.1214/16-ba1018 2017

Showing first 80 references.