Papers

We analyze a variant of stochastic gradient descent with initial regularization (SGDIR) and derive dimension-free upper bounds on its expected excess risk for the squared loss. In the noiseless case, we obtain new bounds for both averaged and non-averaged SGDIR under moment, source, and capacity assumptions. For a particular value of the source parameter, these bounds are of order $m^{-2}\log^{2}m$, where the number of training samples is of order $m$. For another value of the source parameter, we obtain, for any $\epsilon>0$, bounds of order $m^{-3+\epsilon}$, provided that the capacity parameter exceeds $\epsilon^{-1}$. We also establish a lower bound that matches our upper bounds in certain regimes up to a polylogarithmic factor. In the noisy case, we provide an instance-based comparison between SGDIR and ridge regression. Under general assumptions and a mild lower bound on the regularization parameter, we show that the expected excess risk of SGDIR is no larger than that of ridge regression, up to a polylogarithmic factor. Numerical experiments on synthetic and real data are consistent with our theoretical findings.

Data increasingly arrive as collections of curves - a voice recording, a growth trajectory, a day of sensor readings - where each observation is a whole function rather than a single number. The usual question asked of such data is how the average curve differs from one group to the next. But the average is only half the picture: two populations of curves can share almost the same mean and still differ profoundly in how they fluctuate around it, and it is often this variability - the pattern of covariation within a curve - that carries the scientific signal. Comparing populations at this level means comparing their covariance operators, and statistics on covariances is impaired by their non-linearity. The fdWasserstein package equips R users with functions to make such comparisons. It is centered on the geometry of optimal transport, under which covariance operators can be meaningfully averaged, contrasted, and interpolated. It provides the Procrustes-Wasserstein distance between covariance operators, their Frechet mean (barycenter), an ANOVA-type permutation test for the equality of several covariances, principal component analysis of covariance variation, and an entropy-regularized soft clustering of curves by their covariance structure. We outline the underlying ideas, discuss the implementation and demonstrate the complete workflow on the phoneme data shipped with the package.

We develop a finite-width geometric framework describing how learned feature geometries are organized, transported, and selectively aligned in deep neural networks. Incompatibility among weight-generated covariance, gates, and backward sensitivities is quantified through three families of commutators: between gates and covariance, between sensitivities and covariance, and between average gradient outer products (AGOPs) and neural feature matrices (NFMs). An exact layerwise identity decomposes the sensitivity-covariance commutator into four sources: downstream transport, adjacent-layer imbalance, pointwise sensitivity fluctuations, and nonlinear gate-covariance interactions. The AGOP-NFM commutator is a singular-value-weighted transport of the internal commutator, explaining why observed feature-side alignment alone does not determine the internal geometry from which it emerges. Buffered localized energies resolve mixing between separated covariance subspaces. We establish spectral-gap, projector-evolution, and stabilization estimates, and formulate conditional Lyapunov principles that yield decay under explicit geometric error-bound or intrinsic-damping assumptions. These criteria do not follow from gradient flow alone and clarify why risk reduction need not imply commutator collapse. Analytic examples and numerical experiments exhibit factorization of spectral and activation geometry, transient growth, and cancellation among nonzero sources. In tested finite-time regimes, cancellation dominated by a negative transport-imbalance interaction persists across depths, widths, and two regression benchmarks. Spectral alignment therefore appears as a layer- and scale-dependent compatibility phenomenon governed by transport, interaction, cancellation, and possible damping, rather than a universal consequence of training.

We study a resolution problem in local asymptotic decision theory: individual risks may admit Gaussian approximations that do not determine their vanishing difference. A finite menu of experiments is installed before context disclosure, although observations may be routed adaptively afterward. A greatest-element Blackwell order collapses adaptive routing to the best installed experiment and reduces the fixed-menu excess to an inverse-information distortion with frontier $A_k$. We develop differentiated, all-prior posterior transfer along a one-dimensional degradation chain and establish $F_{n,k_n}(H_n)=A_{k_n}\{1+o(1)\}$ for every diverging menu sequence with positive frontier and every admissible localization radius, without an additional direct sample-menu restriction. The transfer is exact under Gaussian degradation. Prior-free likelihood-generator conditions imply it for jump generators and are verified for binary attenuation, Poisson thinning, and negative-binomial thinning. If the distortion is uniformly quadratic on an Ahlfors-regular oracle image of dimension $r$, then $A_k\asymp k^{-2/r}$, and the original-scale excess mean squared error is of order $n^{-1}k^{-2/r}$. Calibrated Poisson sensor and radial-qubit measurement menus illustrate the result. A triangular Gaussian counterexample shows why pointwise Gaussian convergence is insufficient.

External stresses may affect both the circulating levels of specific biomarkers and disturb the correlation structures across molecular entities. The contribution of both types of dysregulations to the subsequent risk of disease are yet to be evaluated. We propose Data Shared Neighbourhood Selection (DSNS), a joint network inference method for estimating preserved and altered conditional association structures across related conditions. DSNS combines neighbourhood selection with the Data Shared Lasso decomposition, representing each nodewise regression coefficient as the sum of a shared component and a sparse condition-specific deviation. This provides an interpretable decomposition of molecular associations while retaining the computational advantages of neighbourhood selection. We also adapt the Stability Approach to Regularisation Selection (StARS) to this two-parameter joint estimation setting. In simulations involving sparse, hub-based and rewiring perturbation mechanisms, DSNS matched the best joint estimation methods for two conditions and outperformed them as the number of conditions increased, while remaining substantially faster than graphical lasso frameworks. Applied to prediagnostic inflammatory proteomic data from future lung cancer cases and matched controls in the EPIC-Italy and NOWAC cohorts, DSNS highlighted altered associations involving CDCP1 and IL10, two established lung cancer risk markers, as well as differential associations involving proteins not selected by risk models.

Analysis of experimental data becomes challenging when the underlying population is connected by a network. Exposure mapping is a common tool in the literature for defining and estimating spillover effects. These mappings reduce the dimensionality of the estimand, thereby facilitating identifiability. It is assumed that this mapping is correctly specified, leaving the choice of the exposure mapping to the analyst. This makes estimators of the spillover effect, such as the Horvitz-Thompson estimator, vulnerable to bias from model misspecification. Although these estimators have been shown to be robust to certain forms of controlled misspecification, there has been relatively little methodological progress in empirically investigating appropriate exposure mappings. In this paper, we propose a novel design-based model specification framework for causal inference. Building on this, we develop a randomization-testing procedure to assess the correct specification of an exposure-mapping model in the presence of network interference. We provide theoretical guarantees for the asymptotic validity of the proposed testing procedure. We establish the favorable power properties of our method through an extensive simulation study and illustrate it in a field experiment investigating the effect of anti-conflict norms among adolescents.

In this paper we propose a new formulation of the Bayesian Filter as used in the discrete-time Markov-Switching-Multifractal (MSM) model of volatility based on existing permutation symmetry within the likelihood structure. We show both analytically and empirically that such a formulation leads to a reduction in time complexity from $O(D^k)$ to $O(k^D)$ thereby significantly reducing the computational bottleneck associated with dimensionality. We compare the agreement between the naive and sector filters and find that while there are significant disagreements, the ground-truth recovery of the latter seems to improve on the former.

Semi-supervised classifiers are commonly trained from samples in which all features are observed but some class labels are missing. When label missingness is independent of the observed data, unavailable class memberships reduce Fisher information relative to a completely classified sample. We study a different regime in which the probability of label missingness depends on posterior classification uncertainty, so that the observed missing-label indicators can themselves carry information about the Bayes decision boundary. Building on the conditionally weighted information decomposition of Ahfock and McLachlan, we develop this phenomenon for a two-component exponential mixture. Although the exponential model is non-Gaussian, asymmetric, and supported on the positive half-line, its log-posterior odds remain linear in the feature. We derive Bayes' rule and its exact error rate, formulate entropy-logistic and squared-discriminant missingness mechanisms, and obtain the full partially classified likelihood. We then derive a decomposition of the Fisher information into the complete-data information, the conditionally weighted loss due to missing labels, and the information contributed by the missing labels. Numerical quadrature identifies regions in which the full likelihood classifier has asymptotic relative efficiency above or below one. Monte Carlo experiments with finite training samples broadly support the population calculations, with the largest departures from the asymptotic predictions occurring near the transition at which the relative efficiency crosses one.

Correlations between rating scales are commonly interpreted as evidence of convergent or discriminant validity, yet prior studies suggest that part of these associations may be attributable to semantic similarity between item wordings rather than to genuine construct overlap alone. Building on this evidence, largely derived from bi-encoders, the present study explores whether cross-encoders, which jointly encode item pairs, offer a suitable technique for detecting semantic overlapping between questionnaire items. Using response data from the NEO-FFI and the PID5BF+M (N = 502, Labek et al., 2024), we examined whether cross-encoder-derived semantic similarity estimates are associated with empirical item correlations, and whether cross-encoders offer a systematic advantage over bi-encoders. Across twelve cross-encoder models, semantic distance was consistently negatively associated with absolute item correlations, reaching statistical significance in two-thirds of the models, with R2 values of up to .37. However, cross-encoders did not consistently outperform bi-encoders based on the same base models. These findings extend prior evidence for semantic components in scale intercorrelations to cross-encoder architectures, while indicating that predictive value depends more on model-specific training characteristics than on encoder architecture itself.

A rapidly growing range of sequential data tasks, such as identifying trend reversals in financial markets, auto-segmenting video and audio recordings, detecting changes in movement direction from motion sensors cannot be fully addressed without detection of distributional shifts in time-ordered data. We consider a sequential change-point detection problem where the conditional density switches at an unknown time, yet neither the pre- nor post-change distribution admits a closed-form. Classical likelihood-ratio statistics are inapplicable in this settings. A conditional diffusion model, trained on pre-change-point data with a frozen context encoder, defines a deterministic bijection via the probability flow ODE. Pre-change observations are mapped onto standard Gaussian latent variables. Post-change observations, processed through the same frozen map, deviate from this reference. We employ the Maximum Mean Discrepancy as the test statistic, derive closed-form expressions for its components under the Gaussian null, and establish its asymptotic distribution as a degenerate U-statistic. Afterwards we apply an online detection procedure of Shiryaev–Roberts to the resulting statistic with exact threshold calibration. The method detects arbitrary distributional shifts, including covariance rotations and higher-order structural breaks, without parametric assumptions on either regime.

Latent trait models differ less in what they measure than in what they fix. Item response theory frees the trait from the items administered but, in doing so, surrenders the origin and unit of the scale to convention. The Cognitive Trait Model (CTM; Choi, 2022) restores both by bounding the trait on [0, 1], where 0 denotes an ignorance level and 1 a mastery level defined by the task domain. We generalize the derivation behind CTM into a family indexed by the mastery anchor $L$: the published model is the case $L = 1$, while the limit $L \to \infty$ gives a half-truncated link with an absolute origin but no ceiling. Origin before unit. Bounded support admits Gauss-Legendre quadrature without truncation error, so item and person parameters follow from Bayes modal EM rather than MCMC. On identical data and priors this reproduces the published estimates ($\theta$ correlation 0.9999) while running 35 times faster than a converged random-walk sampler and 73 times faster than NUTS; a single adaptive-testing update costs about five microseconds. Applied to the time axis, the same link becomes a four-parameter growth curve spanning accelerating, decelerating, plateauing and declining trajectories, though not non-monotone ones. The family reduces to a reparameterization of the logistic model only when all items share a slope, so a measurement advantage is possible in principle; three simulations find none, locating the contribution in interpretation and computation rather than in measurement itself. In the LSAT reanalysis, examinees with perfect scores read 67%, 75% or 100% of mastery depending on the estimator – the criterion-referenced claim is meaningful only with its uncertainty attached, and the coordinate is not in general the proportion of the domain a person can perform.

This paper studies the Sinkhorn distributionally robust hypothesis testing (SDRHT) problem, seeking a robust detector against least-favorable distributions in Sinkhorn discrepancy-based ambiguity sets centered at the empirical distributions. Existing approaches solve this problem by solving large-scale conic programs, which are not scalable. To overcome this, we propose a generative framework that learns least-favorable distributions and supports efficient training and end-to-end sampling. For the Sinkhorn discrepancy-based ambiguity sets, we first derive an equivalent conditional-KL-divergence representation with respect to kernel-smoothed reference distributions. This property allows us to prove strong duality for both constrained and unconstrained minimax SDRHT formulations. Based on the closed-form optimal detector and Brenier's theorem, we reformulate the max-min dual formulation as a maximization problem over convex potentials whose gradients characterize invertible transport maps between kernel-smoothed distributions and their least-favorable counterparts. We efficiently approximate these potentials using Hyper Input Convex Neural Networks (HyCNNs) equipped with stochastic gradient estimators and prove the representation power of HyCNNs and the distributional universality of their induced transport maps. Numerical results show that the proposed method achieves superior accuracy and robustness across different sample sizes and dimensions, while avoiding the scalability limitations of classical SDRHT methods.

Equivalent weights from regression models provide a bridge between design-based and model- based survey inference. We develop a frequentist survey-weighting framework for logistic regres- sion equivalent weights under categorical poststratification. Because the logistic model-based population estimator is nonlinear in the observed outcomes, we define equivalent weights through a local first-order representation, with each weight given by the scaled derivative of the popula- tion estimator with respect to the corresponding observed outcome. This construction yields an explicit closed-form expression in terms of the fitted logistic model and population cell structure. We establish convergence of the effective sample size ratio, asymptotic linearity of the resulting weighted estimator, and a consistent plug-in variance estimator under a finite-cell superpop- ulation framework. Simulation results examine population recovery, regression performance, effective sample size, and the finite-sample behavior of the asymptotic variance approximation. The results show that logistic equivalent weighting provides competitive finite-sample point esti- mation and approximately valid plug-in variance estimation, effective sample size is competitive with alternative model-based weighting in the settings considered. An application to the Fu- ture of Families and Child Wellbeing Study illustrates the construction and use of the proposed weights and variance estimator in an empirical survey setting. We emphasize the relationship between logistic equivalent weighting and recent work on locally equivalent weights for non- linear multilevel regression and poststratification, while focusing specifically on the frequentist interpretation of the weights, their survey-weight properties, and their use for inference.

We test hyperuniformity from one large realization of a stationary point process. Hyperuniformity is equivalent to the average $A_r$ of the structure factor over $B_r$ vanishing as $r\downarrow0$, so the target is a low-frequency average rather than the value of the structure factor at the origin, which need not exist. We estimate $A_r$ from squared Fourier coefficients at frequencies drawn uniformly from $B_r$; if these coefficients are asymptotically Gaussian at almost every fixed frequency, the squared coefficients converge to independent variables with mean $A_r$. Assuming $A_r=s+c r^\alpha+\varepsilon_r$ with $|\varepsilon_r|\leq Lr^\beta$ and given exponents $0<\alpha<\beta$, a two-radius extrapolation removes the $r^\alpha$ term and estimates $s$ with deterministic error of order $r^\beta$. This gives confidence bounds and a one-sided test of $H_0:s=0$ with asymptotic level at most $\gamma$, consistent against every fixed $s>0$ as $R\to\infty$, then the number of sampled frequencies tends to infinity, and then $r\downarrow0$. Essentially free translation actions with completely positive entropy satisfy the Fourier assumption.

Two analysts who calibrate the same predictive model on independent samples will deploy different prediction sets every time, because the calibration threshold inherits the randomness of the data. Wherever deployments must be audited, cached, or approved across sites, this instability is costly: no one can verify that two calibrations produced the same object. We ask two questions: when can independent calibrations yield the identical classifier, and what must that agreement cost? Perfect agreement is impossible, since a procedure that almost always returns one fixed answer cannot remain valid for every distribution, and exact agreement through shared randomness forces the procedure to ignore its data. Sharing a single random seed and rounding the calibrated threshold up to a coarse shared grid resolves the tension: the deployed classifier becomes identical across analysts with any desired probability, coverage guarantees survive, and the price is a quantified increase in set size and calibration data. Matching lower bounds show that no threshold method can pay less, and the method's one tuning constant vanishes asymptotically. Without any shared seed, a fixed grid still confines all analysts to two adjacent classifiers, and no method does better. Replicability also blocks gaming: selecting the most favorable of many recalibrations barely moves a replicable classifier, while the same selection silently undercovers standard conformal prediction. Experiments on real ImageNet outputs, a four-hospital site split, and four language-model families match the theory, including the measured sample-cost frontier.

A rigorous statistical treatment of two-model comparisons on the same evals can be achieved by paired t-tests, analyzing their standard errors and a clustering correction for correlated questions. Nevertheless, leaderboards, ablations studies, and hyperparameter sweeps, usually compare $K>2$ models simultaneously. In this paper, we present a single random-effects model for scores on shared evals. We leverage models as a fixed effects and questions (or question cluster) as random effects. We show that fitting it a classical ANOVA or a linear mixed model we can recover Miller's paired and clustered estimators for K=2, but we extend the results to any K and to unbalanced, clustered designs. We validate the described model in a simulation study and using a real-data application. We take six openly available language models scored on 1,497 shared MMLU-Pro questions across 14 subject clusters, and we show that pairwise ranking claims survive depending on properly accounting for both question-level pairing and multiple comparisons. By the end of the manuscript, we provide concrete recommendations for reporting multi-model eval results.

Q-learning with linear function approximation can be unstable because an arbitrary approximation architecture need not preserve the Bellman contraction. We develop a stable infinite-dimensional linear function approximation framework for Q-learning from a single Markovian behavior-policy trajectory. The learning variable is a coefficient field $\theta\in C(\mathbb L)$ on a compact latent metric space $(\mathbb L,\rho)$. The framework uses a reconstruction operator that maps $\theta$ to a continuous Q-function and a compression operator that maps Bellman updates back to latent coordinates. Nonexpansiveness of both operators induces a contractive latent Bellman map on $C(\mathbb L)$, with a unique fixed point $\theta^*$ whose reconstruction approximates the optimal Q-function up to representation error. We propose two stochastic approximation (SA) algorithms and establish their sup-norm convergence bounds with a leading term of order $\widetilde O(n^{-1/2})$. The infinite-dimensional formulation provides a powerful abstraction for identifying the structures that govern statistical difficulty. Smoothness of the compression map in $\rho$ is inherited by $\theta^*$ and the SA iterates, allowing uniform estimation errors to be controlled through covering numbers of $(\mathbb L,\rho)$ rather than the dimension of $C(\mathbb L)$. Remarkably, the SA algorithms we propose are agnostic to the choice of $\rho$, and thus can automatically adapt to both the smoothness and the geometry. We further illustrate the framework through Q-measure-learning with linear density approximation and output-layer neural weight training under a frozen pretrained network.

Stochastic reaction networks are continuous-time Markov chain models for interacting populations, with applications in biochemistry, epidemiology, ecology, and related areas. We study finite-difference sensitivity estimation when a single estimator requires several nearby parameterized paths. Existing variance-reducing couplings are typically pairwise, so that repeated use is either inefficient or requires application-specific choices in multi-path settings. We introduce the multi-path stacked coupling (MSC), a space-time Poisson construction that jointly generates any finite collection of parameterized paths. Each pairwise marginal of MSC has the same law as the corresponding split coupling pair, allowing existing variance bounds to transfer directly; in finite-state settings, we also obtain first-order expansions for the mean and second moment of finite-difference numerators. We apply MSC in three settings of practical importance: estimating many first derivatives simultaneously, estimating a single first derivative using a wider finite-difference stencil, and estimating higher-order derivatives. Numerical experiments on a processive phosphorylation network demonstrate strong performance in each of the three application areas considered, consistent with the theoretical advantages of MSC: across all three applications, MSC achieves the smallest root mean square error (RMSE) among the methods considered over the tested computational budgets.

During clinical trials evaluating a drug's effect on a survival endpoint, intermediate events often occur in addition to the primary event. The treatment can exert its effect on the primary endpoint along multiple pathways through intermediate events. Assumptions for identifying mediation effects, such as sequential ignorability in natural effects or the dismissible components condition in separable effects, fail because intermediate events act as treatment-induced confounding. To understand the effect along each pathway, we consider hypothetical interventions in transitions between event statuses to mimic the treatment mechanism. The hypothetical interventions adjust for effects through intermediate events and marginalize over unobserved treatment-induced confounding, if any. Based on the derived efficient influence functions for the counterfactual cumulative incidences under hypothetical interventions, we construct multiply robust and semiparametrically efficient estimators for pathway treatment effects. Our proposed framework enables the examination of treatment effects through each transition, on each event, and along each path. By analyzing data from the LEADER Trial, we find that liraglutide significantly reduces the risk of cardiovascular and microvascular events. The reduction in all-cause mortality is primarily mediated by its effects on expanded major adverse cardiovascular events.

Subsampling is effective in tackling computational challenges for massive data with rare events. Overly aggressive subsampling may adversely affect estimation efficiency, and optimal subsampling is essential to mitigate the information loss. However, existing optimal subsampling probabilities depend on data scales, and some scaling transformations may result in inefficient subsamples. This problem is more significant when there are inactive features, because their influence on the subsampling probabilities can be arbitrarily magnified by inappropriate scaling transformations. We tackle this challenge and introduce a scale-invariant optimal subsampling function in the context of sparse models, where inactive features are commonly assumed. Instead of focusing on estimating model parameters, we define an optimal subsampling function to minimize the prediction error, using adaptive lasso to outline the estimation procedure and study its theoretical guarantee. We first introduce the adaptive lasso estimator for rare-events data and establish its oracle properties, thereby validating the use of subsampling. Then we derive a scale-invariant optimal subsampling function that minimizes the prediction error of the inverse probability weighted (IPW) adaptive lasso. Finally, we present an estimator based on the maximum sampled conditional likelihood (MSCL) to further improve the estimation efficiency. We conduct numerical experiments using both simulated and real-world data sets to demonstrate the performance of the proposed methods.

We develop a new framework for flexible, nonlinear, and interpretable off-policy evaluation for infinite-horizon reinforcement learning. To handle large state spaces and support transparent decision-making, we model the Q-function using a nonlinear function class with a sparse additive structure. We derive high-probability finite-sample error bounds for estimating the value function of a target policy and show that the bounds depend only logarithmically on the ambient dimension $d$, thereby alleviating the curse of dimensionality. In contrast to most existing theory for off-policy evaluation, which typically assumes access to many trajectories, our analysis guarantees accurate value estimation when either the number of trajectories or the time horizon is sufficiently large. In addition, we propose a group-sparsity-based feature screening procedure that identifies, with high probability, a reduced feature set containing all relevant covariates. Numerical experiments demonstrate the effectiveness of the proposed approach.

Experimental design under network interference is challenging because outcomes may depend on the treatment assignments of neighboring units. Existing approaches account for network structure but are typically assessed on small or simplified networks, limiting their applicability to complex real-world settings. We propose a network-aware treatment allocation framework that jointly accounts for allocation balance and network topology via an optimality criterion based on the Fisher information matrix. To address the resulting combinatorial optimization problem, we develop an efficient local search algorithm that scales to large networks. We further study the causal properties of the resulting designs by examining the estimation of total, direct, and indirect treatment effects in the presence of interference. Simulation studies across a range of random graph models, including Erdős–Rényi, geometric random graphs, preferential attachment, and stochastic block models, illustrate how network topology influences optimal treatment allocations. Applications to college housing and ego-Facebook networks demonstrate the practical advantages of topology-aware experimental designs.

Model specification tests play a crucial role in evaluating the appropriateness of probability models for estimation and inference. Existing methods for the detection of model misspecification such as the chi-square goodness-of-fit (GOF) tests and more recently the M2 statistic \autocite{MaydeuOlivares2005,MaydeuJoe2014} tend to result in test statistics with excessive degrees of freedom for models with larger numbers of parameters. An alternative approach is based upon the Information Matrix (IM) equality. The IM equality asserts that if a probability model is correctly specified, the asymptotic covariance matrix of the maximum likelihood estimators can be asymptotically estimated using a methodology based upon either the first or second derivatives of the log-likelihood function. Using a contrapositive argument, White (1982) \nocite{Wh82} proposed a misspecification test methodology based upon comparing these two alternative covariance matrix estimators. Extending this work, Presnell and Boos (2004) \nocite{Presnell2004} showed how to develop a misspecification test which only requires one degree of freedom regardless of the complexity of the model or data. In this paper, we extend prior work and additionally apply methods of Golden et al. (2013, 2016) \nocite{golden2013}\nocite{Golden2016} to derive and evaluate misspecification tests for Cognitive Diagnostic Models (CDMs) which only require 1 or 2 degrees of freedom regardless of model or data complexity. Analytic formulas for the tests are derived so they can be applied without requiring computationally intensive bootstrap simulation methods. Our simulation studies show the asymptotic statistical tests have good level (type 1 error) and power performance for CDM models and data which might be encountered in practice.

Theoretical guarantees for kernel regression are typically formulated in terms of smoothness, but practical accuracy depends critically on how the design resolution compares with the target and kernel lengthscales. We develop a finite-sample theory for Matérn regression on periodic domains with quasi-uniform designs, covering noiseless interpolation and noisy kernel ridge regression. Our minimax result shows that accurate recovery requires a design dense enough to resolve the target lengthscale and sufficient information at that scale to overcome noise. We prove that Matérn interpolation additionally requires the design to resolve the kernel lengthscale: if the kernel is too short relative to the point spacing, an additional error remains even when the target is well resolved. For noisy Matérn regression, we derive a three-term fixed-ridge risk characterization consisting of target-scale bias, kernel-scale bias, and variance, and show that optimizing over the ridge parameter yields four distinct contributions. When the target is no more than twice as smooth as the kernel, choosing a kernel lengthscale longer than the target does not worsen the oracle risk, whereas choosing one too short can. Thus these resolution conditions determine when accurate recovery becomes possible, while smoothness determines how rapidly the error decreases thereafter.

The quality of data is crucial for both practice and academia, and this holds for descriptive statistics, AI and advanced analytics alike. This study addresses an apparent gap in the literature and as such presents a full and theoretically grounded overview of data quality assessment approaches and methods, as well as their defining characteristics and inter-relationships. For this purpose a broad typology is introduced that employs two theoretical dimensions, namely the evaluation logic (formal versus informal) and the assessment driver (norms versus data). This yields four high-level approaches and 15 methods for assessing data quality. The research results are relevant for academia, as they provide a theory-based overview and definition of the ways that data quality can be evaluated. The study is also relevant for practice because it allows professionals to make informed decisions on using these methods, e.g. as part of an audit or broad data quality assessment strategy.

The Bandt-Pompe permutation entropy framework, alongside the complexity-entropy causality plane, has become a standard tool for characterizing the dynamical properties of time series. However, observational noise distorts ordinal pattern probability distributions in ways that can systematically misplace time series within the causality plane, compromising dynamical classification. This effect is particularly relevant for geophysical signals, which are typically poorly and irregularly sampled, and have a low signal-to-noise level. In this work, we characterize the distortions on ordinal pattern statistics using the formalism of majorization. We provide theoretical results and propose corrective strategies that restore discriminability under realistic measurement conditions. To achieve this, we introduce methodology that allows the characterization of noisy dynamical series and further allows the quantification of observational noise without the need of a fitting procedure. Finally, we illustrate our methodology by analyzing paleomagnetic records to determine if the geological evolution of the Earth dipole is better described by a stochastic or chaotic system.

We develop a general class of nearest-neighbor non-Gaussian processes (NNnGP) for modeling geostatistical data. By introducing non-linear conditional mean functions into the Vecchia approximation, NNnGP extends the popular nearest-neighbor Gaussian process (NNGP) to effectively capture complex, non-Gaussian spatial characteristics. To ensure coherent spatial predictions from a single realization, we propose a regularized homogeneity condition across the univariate conditionals. We then formulate a Bayesian non-parametric construction of the conditional mean using Gaussian processes and embed the resulting NNnGP as a non-Gaussian spatial prior within a hierarchical regression model. For posterior computation and prediction, we adopt a normalizing flow-based variational inference approach. Numerical experiments on synthetic data and real-world PRISM precipitation anomalies demonstrate that NNnGP captures local non-linear dependencies and significantly mitigates the underestimation of extreme spatial events, all the while maintaining global predictive accuracy.

Bayesian partitioning is utilised simultaneously on subjects and categorical variables to reveal complex dependence structures. Clusters of variables are referred to as views. Variable selection highlights the variables that drive the clustering of the subjects within each view. We derive theoretical results on the relation between the variables' dependence structure and the inferences derived from bi-clustering. The results relate to marginal independence and conditional independence. Using simulated and real data, we demonstrate the applicability of bi-clustering results in assisting log-linear graphical model determination, leading to the efficient exploration of typically vast model spaces. This work sheds light on the relation between two very different but equally popular Bayesian approaches; mixture modelling, that benefits from a large number of variables, and graphical log-linear modelling, which describes explicitly the variables' dependence structure and allows to evaluate the uncertainty associated with model determination.

Near a smooth data manifold, one tangent space summarizes local geometry. At a branch point, the corresponding first-order object is instead a measure over tangent directions, whose normalized masses record the local share of each branch under the chosen data measure. We ask whether a score field at one noise level determines this weighted tangent geometry when the branch center and homogeneity degree $d$ are unknown. In this tangent-measure model, $d$ is the local measure dimension. Gaussian smoothing of a homogeneous tangent measure satisfies an Ornstein–Uhlenbeck eigenfunction equation. Its weak form turns score values—without score derivatives—into a linear system for the center and homogeneity degree, with an explicit rank condition and perturbation bound. After this calibration, the tangential score on one sphere is the spherical log-gradient of a scalar Gaussian–cone transform. Integration recovers that transform up to scale, and all its spherical-harmonic multipliers are positive. Thus one exact shell identifies the normalized angular measure in every ambient dimension $D\geq2$. For at most $K$ positive rays, moments through degree $2K-1$ constructively recover count, directions, and weights in arbitrary dimension. Any fixed observation scheme needs at least $KD-1$ scalar tangential components. In the plane, degree $K$ is both sufficient and necessary, and we give quantitative finite-query certificates. For finite planar $C^{1,\beta}$ branches with positive $C^{0,\beta}$ densities, we prove $O(\sigma^\beta)$ convergence from the finite-noise score to its tangent model. In controlled experiments, 50k-step training lowers validation normalized-score error across four geometries yet raises angular-moment error, separating ordinary score fit from geometry recovery.

Repeated sample splitting, cross-fitting, conformal ensembling, and related randomized workflows often produce multiple dependent valid p-value functions for the same target. Existing p-merging theory guarantees pointwise validity under arbitrary dependence, but turning the merged output into a confidence region typically requires repeated evaluation on a grid over the parameter space. This inversion step can be computationally costly, approximation-dependent, and increasingly difficult to scale. We study when confidence regions can instead be computed exactly from split-wise regions without gridding the parameter space. Our main structural result shows that mergers induced by step calibrators form a broad exactly executable class at the region level, and a partial converse within the calibrator-induced family indicates that exact executability is closely tied to step structure. Within this class, we develop exact voting algorithms, including an adaptive multi-quantile contour aggregator that avoids pre-specifying a single order-statistic threshold while preserving finite-sample validity under arbitrary dependence. Simulation studies on repeated-split regression and conformal prediction, together with real-data regression examples, show that the proposed methods provide stable, robustness-oriented exact inference with substantial runtime gains over grid-inversion comparators. A grid-resolution benchmark shows that fixed-k voting and adaptive multi-quantile voting are essentially insensitive to inversion-grid refinement, while a multidimensional single-threshold stress test illustrates the dimensional blow-up faced by grid-inversion baselines in a simple box-geometry setting. Taken together, the results show that arbitrary-dependence contour merging can be turned into an exactly executable region-computation framework rather than merely a pointwise validity device.