Papers

Heavy-tailed $p$-value combination tests are attractive under unknown dependence because their null tails can be first-order robust even when the exact dependent null distribution is unavailable. That robustness does not resolve calibration: $\Pr\{T>q(\alpha)\}=\alpha+o(\alpha)$ neither quantifies the remaining size error nor determines its sign. We develop a second-order calibration theory for positive Half-Cauchy and reciprocal, or harmonic-mean, aggregation. After exact marginal standardization, extremal dependence is represented by an index-one exponent measure on the coordinate-face lattice. Its support separates axial, full-interior, proper-face, and mixed geometries. A Möbius decomposition shows how the noncompact weighted half-space probes every face, while quantitative face limits determine whether the correction is integrable, critically amplified, or nonintegrable. With common-factor inversion and a tail-to-calibration map, this yields machinery applicable beyond individual parametric copula families. The results include an integrable hidden-face transfer theorem, a sharp fixed-dimensional local-corner theorem with explicit higher-face control, a common-heavy-factor theorem, and size and critical-value expansions relative to independence. Gaussian, standard multivariate-$t$, positive Clayton, and max-linear pair-shock models exhibit distinct power, logarithmic, radial–angular, and proper-face mechanisms; the Gaussian weighted-half-space transfer is conditional on explicit face-boundary hypotheses. Thus dependence geometry determines the rate, coefficient, and direction of the calibration error left unresolved by first-order validity. Asymptotics use $t\to\infty$, equivalently vanishing significance levels.

Forecasting a correlation or covariance matrix is common in risk management and portfolio construction, but evaluating such a forecast correctly is not routine: naive matrix-comparison metrics are not proper scoring rules, walk-forward evaluation windows are easy to overlap with the estimation window in ways that silently leak information, and significance testing on serially dependent forecast-error sequences needs machinery few analysts implement from scratch. corrscore is a Python package that provides matrix-aware implementations of two established proper scoring rules for this setting – the energy score and the variogram score – dispatched across a closed-form tractability spectrum (point, discrete-mixture, and isotropic-Gaussian-mixture forecasts are scored exactly; a general Monte Carlo ensemble falls back to sampling), a geometry-aware variant of the variogram score built from the affine-invariant distance on the correlation manifold, a zero-overlap-by-construction walk-forward backtest harness, and a bundled significance-testing suite (circular block bootstrap, the Diebold-Mariano test, and the Model Confidence Set). We describe the package's design, its point of departure from the existing scoringRules and properscoring packages, and walk through a complete worked example.

Prediction error is widely used to evaluate nuisance-function estimators in causal inference, but its relationship with causal estimator performance may differ across performance measures. We studied this question in a partially linear model using Monte Carlo simulations. We compared ordinary least squares (OLS), generalized additive models (GAMs), XGBoost, and Double Machine Learning with XGBoost (DML-XGBoost), evaluating nuisance-function prediction error, bias, RMSE, and 95% confidence interval coverage. We also examined a simple joint-error measure based on the absolute cross-product of estimation errors from the exposure and outcome nuisance functions. Across the simulated settings, XGBoost had the lowest RMSE among the non-oracle methods, while DML-XGBoost generally provided better confidence interval coverage. Prediction error did not consistently track causal bias across methods and settings, and the method with the best point-estimation performance did not necessarily have the best confidence interval coverage. The joint-error measure was only weakly associated with causal bias and did not provide a useful standalone measure of causal performance. These results suggest that prediction error is useful for assessing nuisance-function estimation, but it should not be treated as a direct measure of the quality of the resulting causal estimator.

We introduce a robust clustering method, MK-means DPD, that estimates cluster centers and covariance matrices using density power divergence (DPD) measures combined with Mahalanobis distance, making it resistant to outliers and adaptable to heterogeneous, elliptical clusters, unlike the classical K-means algorithm. Since Mahalanobis distance-based K-means lacks a general convergence guarantee, we further introduce a convergent variant, Density-Consistent MK-means DPD (DC-MK-means DPD), which redefines the cluster assignment step in terms of a pointwise DPD loss. We prove a formal theorem establishing that the resulting algorithm converges in a finite number of steps. We also propose two new robust internal evaluation indices, a Median Davies-Bouldin Index and a Trimmed Calinski-Harabasz Index, to ensure that performance comparisons are not themselves distorted by outliers. The efficacy of the proposed methods is demonstrated on simulated data, showing superiority over existing methods, and on two real datasets: Iris data, to identify similar species, and COVID-19 case fatality rate and infection rate data for countries worldwide, examining the resulting clusters' geographic and socio-economic patterns.

We introduce a distribution-free goodness-of-fit test, termed the omega-1 test, which naturally complements the Kolmogorov–Smirnov test and Cramér–von Mises test and can be viewed as their (piecewise) linear analog. Defined as an $\mathrm{L}^{1}$-functional of the empirical process, the test statistic improves on balancing sensitivity to localized and diffuse alternatives and gives a robust and interpretable measure of distributional discrepancy, apart from close connections to the Wasserstein 1-distance. For finite samples, we derive a finite-dimensional computational form for the statistic under general conditions, which leads to various explicit formulas for its null distribution. Under mild continuity assumptions, the limiting statistic is distribution-free, with explicit distribution formulas. In composite settings, the statistic is also compatible with the Khmaladze transformation, enabling asymptotically distribution-free testing. The limiting transformed statistic also has an explicit distribution that escapes reliance on intractable compensator processes or purely numerical evaluation. Simulation results indicate rapid convergence of the finite-sample distributions to their limiting counterparts and support the practical applicability of the test.

Token prediction is a central pre-training objective for modern language models. Despite its empirical success, why token prediction learns broadly useful representations remains incompletely understood. We develop a statistical framework connecting token prediction with representation geometry, encoder approximation, and downstream performance. Under a softmax prediction head, we show that accurate token prediction organizes token embeddings according to similarities between the distributions of contexts in which different token types appear, as measured by Hellinger distance, with explicit errors governed by prediction accuracy and token frequency. Meanwhile, the contextual representation provides a low-dimensional coordinate for the conditional distribution of the target token relative to these embeddings. We further introduce a self-consistency principle showing that repeated applications of a shared representation block can progressively refine the contextual representation without introducing additional block parameters. Among representations with the same prediction accuracy, this recurrent construction favors those that can be stably reconstructed from their contexts. Finally, we establish downstream guarantees for token generation, token community recovery, and classification by a linear probe, showing how prediction accuracy and recovered geometry translate into performance beyond the pre-training objective. Together, these results explain how the simple objective of predicting tokens can recover semantic geometry and produce broadly useful representations. A controlled simulation illustrates the theoretical mechanisms.

Missing data, measurement error, and population heterogeneity are pervasive challenges in analyzing data arising from modern observational studies and machine learning applications. Although these problems frequently coexist and interact, they are often treated separately in existing works. We propose a unified probabilistic framework that jointly addresses these issues utilizing deep latent variable representation. The proposed method integrates a novel hierarchical tree-routed variational autoencoder with pattern-aware latent representations and calibration-based denoising. The framework accommodates missing data mechanisms, including MCAR, MAR, and MNAR, while simultaneously learning subgroup-specific and globally shared latent structure. The introduced reconvergent routing mechanism enables selective parameters to be shared across related subpopulations, which offers flexibility as well as improved statistical efficiency. Simulation studies demonstrate substantial improvements over existing deep generative imputation approaches under complex heterogeneous missingness and measurement-error settings. The proposed framework provides a principled approach for learning from noisy and incomplete data in modern healthcare and other high-dimensional applications.

Integrals of spectral densities are frequently used to summarize spectral characteristics of linear processes. This work studies self-normalization for estimators of such integrals based on sequential periodograms and establishes weak convergence of the corresponding processes. For linear functionals of the spectral density, self-normalization yields pivotal limiting distributions that are free of unknown spectral quantities. For non-linear functionals, however, additional components with distinct covariance structures may arise in the limiting process. We demonstrate this phenomenon for the integrated squared spectral density.

Introduction: Longitudinal lung function recovery after pulmonary disease commonly follows nonlinear trajectories, and failure to adequately model these trajectories can lead to biased or misleading estimates of treatment effects. However, an important methodological gap remains as there is limited assessment of statistical methods for modeling nonlinear lung function trajectories. Methods: We compared several longitudinal modeling approaches for characterizing recovery in percent predicted forced expiratory volume in one second (FEV1p) using data from a phase 2 randomized trial for pulmonary tuberculosis (TB). We estimate the differences in repeated mean FEV1p between each treatment arm and control arm over a 180-days follow-up period. We compared 6 different statistical models: (1) linear mixed-effects model, (2) a piecewise linear mixed-effects model, (3) quadratic and (4) natural cubic spline mixed-effects models, (5) a nonlinear mixed-effects model with exponential recovery function, and (6) a generalized additive mixed model. We discussed the assumptions, clinical interpretations, and resulting treatment-effect estimates across these approaches. Results: The results from the TB trial analyses showed that the conventional linear mixed-effects model provided limited evidence of treatment differences over follow-up, whereas several flexible models identified significant differences during specific periods of recovery. Conclusion: Flexible longitudinal models can complement conventional linear mixed-effects models by revealing treatment differences at certain periods of follow-up that may be obscured by assuming a single linear trend. The choice of nonlinear modeling strategy should be guided by the scientific objective, available data, and the desired balance between clinical interpretability and flexibility.

Compositional data are positive multivariate data whose sum equals 1. A popular method to analyze such data is via log–ratio transformations, which are however not applicable when zero values are present. In this paper we present a conditional logistic normal distribution suitable for compositional data with structural zero values. The model is applicable to arbitrary dimensions and the EM algorithm guarantees a fast implementation. The model accepts predictors and we provide a residual analysis. We further provide principal component analysis and biplots for visualization of compositional data with zero values present. Simulation studies and real data examples illustrate the performance of the model, with and without covariates, and compare it to the Dirichlet analogue.

Via Computational Taxonomy (CT), we develop Design of Experiment(DoE) based on rigorously redefined constituting ingredients of complex system dynamics: randomness, nonlinearity and even class, through a data-driven constructed Taxonomic Hierarchy. As an opposite quest of Classification without man-made assumptions and structures, we illustrate this new perspective of DoE through a civil engineering complex system: Concrete Compressive Strength (CCS). CT begins by building a Taxonomic Hierarchy as a heterogeneity-vs-homogeneity map framed with a tree geometry to represent CCS-system dynamics. At each internal node of this hierarchy, a heatmap is computed via Scientific Data Analysis (SDA) to reveal locality-embraced heterogeneity through block structured covariate homogeneity annotated with response's locality-split. Only arriving at each ending-node of this hierarchy, coherence of homogeneity is achieved on both response and covariate sides. As such a class of finite sample nature is computationally recognized and confirmed. In contrast, nonlinearity is evidently observed as incoherence of response-vs-covariate homogeneity when comparing two classes located on two distinct branches. This hierarchy explicitly maps out system's randomness and nonlinearity to serve as a scientific basis for any DoE quest. Design of Experiment (DoE) for any designated class is redefined as a search for a covariate subspace that embraces the class-representative randomness and at the same time avoids potential nonlinearities with respect to the rest of classes. This is a brand-new theme of DoE.

This paper studies the problem of model order selection in stationary noise. Single-threshold detection is sensitive to extreme noise excursions, particularly when the detection threshold is lowered to capture weak deterministic components. To augment single-threshold detection, we propose a differential step-down thresholding algorithm. We use a threshold grid in this algorithm. To overcome the threshold grid misalignment error induced by grid evaluation, we utilize randomized grid dithering. Using extreme value theory, we show the clustering of the noise extrema. The stopping rule of the proposed algorithm detects this clustering and stops the algorithm to prevent false alarms. By analytically bounding the threshold grid misalignment error, we prove that our algorithm achieves asymptotic exact order recovery under the 0-1 loss function. Moreover, the proposed algorithm remains robust even if the detection threshold in the original single-threshold algorithm is lowered or extreme noise excursions occur.

Multilinear principal component analysis (MPCA) reduces the dimension of tensor-valued data while preserving their mode-specific structure, but its quadratic scatter criterion can be unstable under heavy-tailed distributions and contamination. We propose spatial-sign-based multilinear principal component analysis (SMPCA), a robust dimension-reduction method that centers the observations by their spatial median, removes radial magnitude through spatial-sign normalization, and estimates the mode-wise loading spaces by alternating eigendecompositions. Under a separable tensor elliptical model, we show that the target mode-wise loading spaces uniquely maximize the population criterion and that one complete sweep of exact population block updates recovers them from any initialization. We also characterize exactly when their tensor-product subspace coincides with a leading unrestricted subspace of vectorized spatial-sign PCA and, when finite second moments exist, ordinary vectorized PCA. At the sample level, we derive explicit statistical rates for the mode-wise subspaces and the joint multilinear projector, obtain corresponding reconstruction guarantees, establish consistency of the cumulative-contribution dimension selector, and prove that the objective values generated by exact cyclic updates are nondecreasing and convergent. Simulations and an empirical application show that SMPCA is more accurate and stable than competitors under heavy-tailed distributions and outlier contamination, while retaining competitive performance under light-tailed settings.

Win statistics require a prespecified outcome hierarchy. Clinical teams sometimes agree only on the highest priority outcome, leaving the order of lower priority outcomes unresolved, and clinically meaningful thresholds may be specified as ranges. Separate sensitivity analyses describe how the results change. A single inference for the full set of planned analyses is generally absent. We define the estimand as the smallest net benefit among the comparison rules allowed by the protocol or statistical analysis plan. Each rule uses the usual two-sample U-statistic. Large-sample results are developed for a finite list of rules and for a continuous threshold range. Inversion of an intersection-union test gives a one-sided lower confidence bound, with no multiplicity adjustment for the single claim that every individual net benefit is positive. Simulations show valid one-sided coverage and illustrate the gap between a favorable result for one selected hierarchy and a favorable result across all prespecified hierarchies. An application to ACTG 175 shows how the ordering of laboratory outcomes can affect the strength of the conclusion. A further example identifies an unfavorable threshold that is missed by a sparse grid.

We derive the complete duality theory underlying the Hodrick–Prescott filter, whose rank-deficient second-difference penalty admits infinitely many equivalent trend representations via generalized inverses. Constructing two canonical choices—the Moore–Penrose-based B-representation and an alternative A-representation—we prove that the extracted trend is invariant across representations, obtain a closed-form Bregman-type divergence quantifying their disagreement under a shared coefficient vector, and show this divergence vanishes as $\lambda\to\infty$. We further mollify the non-smooth $\ell_1$ trend filter with a compactly supported biweight kernel to obtain a closed-form $C^3$ Convolutional LASSO that restores Newton-type quadratic convergence without sacrificing the kink- setecting character of $\ell_1$ regularization. Theoretical results are verified numerically and benchmarked against ADMM/IRLS on NVIDIA's 2013–2018 daily closing prices, where the sparse filter isolates genuine growth-regime breaks—chiefly the November 2016 post-earnings acceleration—cleanly separated from the smooth $L_2$ trend.

Renewables now dominate Great Britain's generation mix but rarely occupy the marginal price-setting position, which raises the question of which flexible technologies translate a renewable-rich system into real-time price formation. This study reconstructs the price-ranked edge of the eligible bid or offer stack in the GB Balancing Mechanism across 50,684 half-hourly Settlement Periods from 2023 to 2025 and attributes it to individual Balancing Mechanism Units, separating long-system bid-active from short-system offer-active conditions and retaining co-marginal ties. Batteries rose from 0.8% to 36.2% of bid-active and from 3.2% to 26.9% of offer-active marginality, displacing combined-cycle gas on the bid side and pumped storage on the offer side. Capacity-normalised marginal capture rose on both sides, and unit-quarter fixed effects show that an additional 100 MW of Balancing Mechanism active capacity was associated with a 0.55 and 0.48 percentage-point increase in quarterly bid- and offer-side marginal share. By 2025, battery actions were around {\pounds}9 to 10/MWh more favourable than volume-matched non-battery alternatives drawn from the same ladder, yet batteries were under-represented by 12.9 percentage points in the highest-priced 5% of short-system periods. Individual batteries remain consistent with price-taking, while the fleet has become endogenous to routine balancing price formation.

We study the eigen-structure of the penalty matrix arising from cubic smoothing splines on equally spaced knots. Using purely matrix-theoretic arguments, we show that its positive eigenvalues are simple, that the associated eigenvectors alternate between even and odd, and that the eigenvector for the $k$th largest eigenvalue has exactly $k+1$ sign changes. The approach provides a direct and transparent alternative to existing variational proofs of the oscillation property. These results show that equally spaced knots support a spline basis with both a parity structure and an oscillation pattern.

Understanding why a target metric changes is a fundamental problem in data-driven decision making, beyond anomaly detection alone. We study root cause attribution for metric changes in complex e-commerce systems, focusing on trade-offs between interpretability, efficiency, and causal validity. As a starting point, we extend a metric-decomposition method into a recursive metric-tree framework for multi-level root cause analysis, but this relies on independence and decomposability assumptions that miss complex causal dependencies. In contrast, graphical causal models (GCMs) relax these assumptions and improve causal validity, at the cost of interpretability, higher computational and data demands, and potential attribution target misalignment. Through real-world applications, mathematical proofs, and simulations, we characterize the fundamental sources of these trade-offs. Guided by these insights, we propose a unified, causally informed attribution approach that integrates structural causal information into the metric-tree decomposition framework and corrects key sources of misalignment in GCM-based causal attributions, substantially improving causal validity while preserving interpretability and fast computation. Analytical proofs and simulations demonstrate that the proposed approach produces more accurate root cause attributions, and we also present a real-world application.

Longitudinal cohort studies produce repeated data that enable the assessment of time-varying association patterns between exposures and health outcomes. Classical linear mixed-effects models (LMMs) can accommodate a large variety of association patterns while accounting for the irregularly spaced, partially observed measurement. But they require the analyst to pre-specify the functional form linking the exposure history to the outcome. We propose the Neural ODE-LMM, which embeds a Neural Ordinary Differential Equation (Neural ODE) within the linear mixed-effects framework: a learned vector field encodes covariate trajectories into a continuous-time latent state that drives both the fixed- and random-effect design, while preserving the standard LMM observation model. This retains classical likelihood-based inference while learning complex, potentially cumulative, covariate effects flexibly. All parameters are estimated by maximising a penalised marginal likelihood. To quantify covariate effects, we introduce contrasts of counterfactual predictions that compare the expected outcome under alternative covariate trajectories with variance estimated via the delta method. In simulations, the model recovers both instantaneous and cumulative-burden effects without prior specification of the functional form. Applied to the Trois-Cités (3C) cohort, a population-based study of 7{,}324 participants, the method reveals trajectory-dependent associations of BMI and fasting glucose with cognitive decline.

The business of life insurance companies is characterized by long-term contracts. For this reason, data describing customers is of immense value. A portion of the data provided to the customer is rarely or not at all analyzed. This includes customer letters of any kind. This work focuses on classifying customer letters as cancellations and identifying the respective reason, if available. The outlined approach can also be applied to other business transactions and reasons. We discuss data acquisition and preparation, present alternatives, and explain the reasons for the chosen approach. A successful implementation of such a tool can lead to a better understanding of customer cancellation behavior by the insurer, enabling more targeted actions in certain situations.

The Antarctic ice sheet is a major source of uncertainty in future sea-level projections, and physical simulators such as the PSU3D-ICE model are essential for studying its evolution. Calibrating them against observations is challenging: the simulator outputs and observed ice-thickness fields are high-dimensional, spatially dependent, and semi-continuous, with a large point mass at zero denoting ice-free regions. These features make conventional Gaussian-process emulation and likelihood-based calibration ill-suited and computationally infeasible at full resolution. We propose the sequential calibration method guided by a diffusion model and a Siamese network (SC-DS), a fully neural framework. For emulation, we develop a single conditional diffusion model that jointly generates the binary ice presence–absence pattern and the continuous thickness field, using global and local conditioning to represent how the input parameters shape the output. Because the emulator induces an intractable likelihood, our likelihood-free calibration replaces the hand-chosen distance and tolerance of approximate Bayesian computation with a probabilistic acceptance rule learned by an iteratively retrained Siamese network, together with a data–model discrepancy adjustment. Applied to the West Antarctic Ice Sheet, SC-DS matches the accuracy of state-of-the-art Gaussian-process calibration at a fraction of its computational cost and scales to the full-resolution domain, where existing methods become intractable.

Statisticians have spent over a century arguing that we are more than calculators, usually by pointing to what else we know. AI is making that defense harder, since the list of what only statisticians can do grows shorter with each model release. AI makes claims cheap to generate and may eventually make the statistics behind them cheap too. However, a model cannot be answerable in the way that statistical practice requires. Statistical leadership then becomes a question of which claims we are there to answer for, including the ones we answer for in advance by building judgment into systems.

In this article, we revisit the problem of group factor analysis and propose a one-step penalized least squares method to estimate the factor loadings and factors in large-dimensional group factor models, offering a distinct alternative to the conventional two-step principal component approach. Our procedure originates from the equivalence between the group factor structure and the carefully tailored identification conditions. Leveraging this insight, we develop a tricky Lagrange multiplier formulation for a penalized least square loss function. This one-step optimization framework, combined with the fine-tuned penalty parameters, facilitates the direct derivation of central limit theorems for factor loadings, factor scores, common and local components, as well as the convergence rates for them. Our theory demonstrates that our one-step approach achieves the same rate as the two-step aggregated principal-component method and even the same limiting standard error in the balanced group panel case, but smaller limiting standard error than the two-step canonical correlation procedure. Extensive simulation studies justify the theory. Applications to U.S. house prices and CSI300 weekly returns confirm that our method identifies global factors and heterogeneous local patterns.

Functional clustering is an important tool for identifying latent heterogeneity in functional data and has been widely applied across various scientific fields. However, many existing methods are not fully adaptive, as they may require the number of clusters to be prespecified and may lack automatic control over the smoothness of the underlying functions. They also commonly assume independent and identically distributed errors, thereby overlooking additional within-curve dependence. We propose a fully adaptive Bayesian procedure for functional clustering that addresses these limitations through Dirichlet process priors, adaptive smoothness control, and flexible covariance modeling. For computational scalability, the proposed method employs variational inference as an efficient alternative to Markov chain Monte Carlo. Together, these features provide a unified Bayesian framework for functional clustering and cluster-specific mean function estimation in the presence of structured within-curve dependence.

A continuous clinical measurement recorded in fixed physical units may have different prognostic meaning across patients when its effect depends on a patient-level modifier. For example, the same tumor diameter may imply markedly different prognosis in an infant and an adult, because patient size can modify its prognostic effect. We formalize this problem through prognosis-equivalent mapping for time-to-event outcomes. The resulting estimand maps a measurement value observed at one modifier level to the value under a reference modifier level that yields the same conditional prognostic quantity. The basic operation is to equate a monotone conditional prognostic score and invert the reference-side curve. In observational data, however, treatment may be selected according to the modifier and the measurement, so a direct equate-and-invert procedure may reflect differences in treatment assignment as well as the prognostic meaning of the measurement itself. We therefore define, in addition to a direct approach, a policy-standardized mapping that standardizes treatment under a common policy using the g-formula. We also introduce an origin-referenced mapping that compares changes in prognostic score from a common anchor, thereby separating modifier-specific prognostic levels from modifier dependence in the measurement–prognosis relationship. Using Cox models, we develop linear and flexible inversion estimators with analytic and bootstrap confidence bands. Simulations evaluate finite-sample performance and illustrate how treatment standardization and origin referencing clarify what is being mapped.

The main focuses of the article are limit theorems of Wasserstein projection in the convex order which are useful for inference tasks. The main results rely on the establishment of dual attainment, stability and several useful observations. The first is a clean criterion to invoke the Wasserstein projection dualities on the classical cost $c=h(x-y)$. A new strategy was introduced to establish the dual attainment. Backward and forward dual transport are proved to enjoy a conjugate relationship. Sufficient conditions are given that ensure uniqueness of the optimal dual potential. These ingredients laid the groundwork for the proof of optimal dual potential stability, thereby permitting us to prove the central limit theorems under mild moment assumptions. Additional discussions show that the moment assumptions are sharp. These results also answered several open questions raised by Professor Benjamin Jourdain.

A company with a fixed artificial intelligence (AI) budget must decide which large language model (LLM) handles each recurring workload. What it lacks is the quality table, how well each model performs on each workload. Given that table, the decision is a multiple-choice knapsack problem and is routine to solve, so estimating it is the difficulty, and that estimation fails in two ways. Models are rarely compared on the same work, and the recorded score is usually a proxy rather than the outcome the company values. Causal and off-policy methods repair the first but condition on the second, while evaluator-validation methods estimate the second but stop short of the decision. Worse, buying more re-evaluation cannot settle the second: randomization governs which requests are scored, not how a score is produced, so the table stays uncertain however much evaluation is purchased. Yet the deployment decision may still be determined even when the table is not. We therefore ask whether one assignment stays optimal across every quality table consistent with the evidence. For the fixed-budget problem, this admits an exact two-solve certificate: solve once at the estimated table and once at a least-favourable table. Agreement certifies the assignment; disagreement identifies the model-workload pairs where further evidence can matter. We propose CASE (causal active sequential experimentation), which targets evaluation to those pairs and repeats the test as evidence accumulates. On a production log, the measurement failure is the larger of the two: correcting assignment exactly still leaves most of the loss, and randomized re-evaluation does not remove it. In our experiments, the available evidence often does not determine the assignment. On paid software tasks, better information about model quality yields more savings than further optimization of the assignment on the same estimates.

In resin transfer moulding, complete saturation of the fibre preform is necessary before the resin front reaches the outlet vent(s), to prevent dry-spot formation. In practice, the flow front rarely advances uniformly due to race-tracking effects. We propose a combined estimation and control strategy to address this issue. We use pressure-sensor data collected during filling to estimate the unknown race-tracking strengths via an iterated extended Kalman filter, and to simultaneously optimise auxiliary gate pressures to prevent the resin from arriving at the vent before complete saturation occurs. To make this approach feasible in real time, we replace the expensive finite element-control volume model with neural network surrogate models. The Bayesian approximation error framework is used to account for the discrepancy between the surrogates and the original model. To make the training of the surrogates feasible with time-dependent control actions, we prove that the flow-front geometry depends only on the time-averaged gate pressures. We demonstrate the methodology on a representative forked geometry, showing that the combined estimation-and-control framework can substantially reduce the number of unfilled nodes at the end of filling compared with an uncontrolled baseline. However, the level of improvement depends strongly on the location and number of auxiliary gates.

Although Pollak's minimax formulation is one of the central frameworks in quickest change detection (QCD), the strongest general optimality results available for it are predominantly asymptotic, applying as the average run-length-to-false-alarm constraint, $\gamma$, tends to infinity. Exact results for finite $\gamma$ have previously been available only for special models or restricted regimes. This paper addresses the general finite-$\gamma$ problem over the complete class of randomized, history-dependent stopping rules for i.i.d. change models. The key technical development is a survival-process representation that recasts the optimization of Pollak's minimax criterion over stopping times as an equivalent linear variational optimization over admissible survival processes. Although this formulation is infinite-dimensional, it establishes the existence of an optimizer and provides an exact characterization of the finite-$\gamma$ Pollak minimax value. This characterization, in turn, provides a principled basis for constructing computable stopping rules whose performance can be made arbitrarily close to the optimum. The resulting rules are driven by a recursively updated weighted likelihood-ratio statistic with generally time-varying injections and boundaries. Importantly, this structure is not imposed a priori: the optimization is carried out over the full class of randomized, history-dependent stopping rules, and the Shiryaev–Roberts form emerges naturally from the solution. In particular, the classical Shiryaev–Roberts recursion arises as the time-homogeneous special case. Finally, under a likelihood-ratio floor condition, the framework yields closed-form exact minimax solutions for a nontrivial class of change models.

Worst-case Optimal Recovery (OR) and Bayesian learning describe the same Gaussian-quadratic-Hilbert problems in two vocabularies. We sharpen the correspondence - the radius of information equals a nugget-optimized GP posterior variance and is attained by the posterior mean at a closed-form balance nugget - and measure, inside three published Bayesian systems, where the worst-case side pays. The ledger is two-sided: the losses instruct as much as the wins. Morozov calibration tracks a test-access oracle within $1.00$-$1.19\times$ where $\sigma$-blind rules fail, is $4.9$-$6.3\times$ more reproducible across noise draws ($p=0.002$-$0.004$), and is the only deployable rule whose selection survives a change of backend ($1.36\times$ against $12$-$30\times$ for the released weight, ML-II and GCV); tight certificates cover at the information-theoretic floor with no numerical slack. But on exchangeable data split-conformal beats the OR head on interval score, a water-filling prior adds nothing without an oracle noise hint, and under covariate shift the OR band keeps coverage on every dataset yet loses interval score to split-conformal, and to a feature-free constant band, on most cells; what pays is not shift but shift on a learnable target, which a training-free audit statistic predicts before any model is fitted. In Bayesian optimization the certified width is a validity floor whose scalar inflation we prove inert under a checkable margin condition and check at every step. Inertness is graded, not binary, and in the size of the inflation as much as in the objective: $\kappa{=}2$ is inert wherever $\kappa{=}5$ is and on more cells besides, while $\kappa{=}5$ moves half the Ackley seeds and every Griewank seed. Exploration is a shape problem, not a scale one. The design rule: match the guarantee tool to the data regime, and audit the regime first.