Stat arXiv of the Last Week — 29 Oct–5 Nov 2026

Introduces Inherited Local Shrinkage, a Bayesian global‑local shrinkage framework for adaptive functional form discovery via DAG‑based hierarchical shrinkage.

Proposes group sequential ADDIS-Spending algorithm to control FWER online in multi-stage platform trials.

The paper introduces a Spider Network Filtering Framework to encode interactions for logistic decision-making in bank acquisition analysis.

Introduces a Bayesian latent space hidden Markov model to cluster longitudinal multilayer networks with regime-specific latent configurations.

Provides an exact Poisson mixture representation for Siler remaining life expectancy.

Unifies various reweighting schemes for stratified path sampling, showing they share common underlying estimator.

Introduces a tractable state‑space chain‑ladder model with exponential seniority weighting and closed‑form uncertainty calculations.

Critiques predictive-fit metrics for marketing mix models and proposes using experimental benchmarks for model selection.

Estimates time-varying extreme value index using a ratio‑form kernel Hill estimator and constructs confidence bands for risk measures.

PFNs estimate posterior inclusion probabilities for variable selection in Bayesian networks without posterior sampling.

The paper evaluates clinical trial simulators' interventional validity using an intervention-blind test on thousands of trials.

Develops a competing risks framework to model correct, incorrect, and unfinished token completion under budget constraints.

Proposes doubly robust win statistics for hierarchical survival endpoints in individual and cluster-randomized trials.

We use conformal prediction to generate diagnosis sets from incomplete Parkinsonian clinical records.

The paper introduces a regularized local Whittle estimator for long‑memory parameters in density time series via Bayes Hilbert space.

Uses exponential random graph models to analyze influence ties among jazz trumpet players.

Introduces Balance-Guided Forest to jointly learn covariate balancing and tree-based representation for causal effect estimation.

Provides a unified approximation framework for inventory and assortment optimization under general choice models.

Develops a Vecchia‑based Gaussian process framework for scalable analysis of large replicated and censored data.

Implement four-level structured functional PCA for unbalanced hierarchical data using memory‑efficient covariance estimators.

Develops Bayesian superpopulation models to estimate diabetes prevalence from multi‑site EHR data, evaluating ANOVA, geographic, and MRP approaches.

The paper evaluates simple household fuel purchase strategies across different price volatility models.

Learns componentwise transport maps via annealing paths to improve multimodal sampling and Bayesian inference efficiency.

Proves geometric ergodicity of a three‑block Gibbs sampler for Bayesian linear mixed models with scale‑mixture random effects.

Presents a non‑parametric kernel estimator of mixed covariance density between point process and random field for dependence testing.

Develops a regime‑aware survival estimator that recovers true effects when post‑progression treatment schedules are provided.

The paper evaluates facility proxy strata for variance estimation and confounding adjustment in claims data without identifiers.

Product Monte Carlo estimates expectations under a product of densities using only factor samples, via regression and Bayesian uncertainty.

Adapts reliability engineering methods to define AI system failures and proposes a diagnostic framework with statistical measurement guidance.

Develops optimal distribution‑free sequential changepoint detectors using conformal martingales with minimax‑optimal delay bounds.

Develops an EM algorithm for maximum likelihood estimation and imputation of missing values in wrapped elliptically symmetric torus data.

Introduces curvature-calibrated quasi-Bayesian updating to improve moment‑restricted posterior approximation.

Provides a fully conjugate Bayesian hierarchical framework for covariate selection in exponential family GLMs with posterior consistency and Gibbs sampling.

Provides a model‑agnostic statistical framework to audit privacy disclosures in synthetic data via hypothesis testing.

Empirically compares 56 stochastic optimizers for tuning‑free variational inference across diverse Bayesian problems.

Anchor PCA modifies PCA target matrix to obtain shared low-rank embeddings across domains.

The paper establishes that predictors based on invariant stable blankets achieve worst-case optimal risk in prediction-intervention games.

Demonstrates that any distribution‑free prediction set for ITEs necessarily has infinite expected length.

Derive first-order bias bounds for stochastic gradient MCMC invariant measures using Gaussian convolution inequalities.

Provides exact parametric tests and confidence intervals for type‑I censored Weibull reliability data.

Introduces a scalable generalized deep Gaussian process framework for emulating simulators with heteroskedastic and non‑Gaussian outputs.

The paper proposes a hybrid GAT and latent Gaussian spatial process for spatio-temporal risk mapping.

Develops weighted clustered U‑statistic estimators for generalized causal effects across multiple outcomes in cluster‑randomized trials.

Introduces a goal-oriented sequential Bayesian experimental design using variational inference and amortized policy networks for causal quantity estimation.

Recurrent networks automatically learn changepoint penalty parameters from raw sequences.

RCOMPSs provides a scalable task-based runtime for parallel execution of R code on multicore and manycore systems.

Introduces a global‑local Dirichlet‑process clustering framework for high‑dimensional function coefficients with spatially correlated errors, implemented via MCMC.

The paper introduces a label‑free churn prediction score based on six‑step propagation of ergodic Markov chains in social networks.

Introduces a shrinkage-prior based generalized Bayes method with block SGMCMC for scalable inference in high-dimensional implicit generative models.

Introduces post-stratified importance sampling using Gaussian process regression and heavy‑tailed density estimation to improve variance and uncertainty quantification.

Introduces an exchangeable particle Gibbs sampler for discretised Markov jump processes, improving mixing efficiency for Bayesian inference on reaction networks.

Proposes a model-based bootstrap to improve end-to-end digital twin prediction accuracy and coverage.

Introduces a Hidden Markov Ising model with exogenous covariates, using pseudolikelihood for scalable temporal network inference.

Derives exact distributions for prior‑informed stepwise SNP selection in GWAS.

The paper reviews multifidelity techniques for scalable risk and reliability analysis using control variates, importance sampling, and active learning.

MambaPFN pretrains a Transformer to amortically select nonlinear variables using posterior inclusion probabilities.

Derives cluster‑adjusted standardized mean differences from Bayesian multilevel location‑scale models via partial pooling.

The paper introduces bootstrap re‑selection procedures to quantify coefficient uncertainty after Lasso‑selected Cox model fitting.

Develops a Bayesian changepoint detection method with a decision score to prioritize district heating faults.

Presents a geometry‑driven framework for directional statistics and related estimation methods in signal processing.

Develops a Bayesian Gamma-based method to aggregate incomplete reliability data for fusion devices.

The study applies competing risks Cox models to analyze soccer coach turnover in Serie A.

Derives detection limits for finite versus infinite population size via profiling and likelihood ratio analysis.

Derives exact MISE-based and smoothed cross‑validation bandwidth selectors for Dirichlet kernel density estimation on the simplex.

The paper outlines conditions under which complete-case regression yields consistent causal effect estimates using DAGs and weighting.

The paper introduces calibrated upper L‑statistics for high‑dimensional mean change‑point testing with adaptive rank aggregation.

Provides a model-assisted covariate-adjusted estimator for time‑restricted marginal win statistics with guaranteed asymptotic efficiency.

The paper reviews recent Bayesian small area estimation methods, including EBLUP and hierarchical Bayes, for area and unit level parameters.

Introduces a Bezier‑surface Gaussian Process to model wafer‑level chip warpage using low‑dimensional control points.

Reviews how machine learning and uncertainty quantification integrate into real-time decision-making pipelines for digital twins.

Provide a debiased one-step estimator for physics parameters via semiparametric influence functions.

Evaluates bounded-memory calibration effects on forecast interval coverage during recurring high‑volatility states.

Introduces deterministic uniformization EM algorithm for interval‑censored multistate transition model estimation and latent trajectory inference.

Introduces an iterated neural estimator for nonparametric GARCH modeling with finite-sample error bounds.

Estimates canopy cover fraction using inverse-probability weighting and validates with a Cramér‑von Mises test.

Provides minimax detection rates for regression and GLMs under differential privacy.

Formulates weighted, regularized isotonic regression on graphs as a sparse generalized lasso with a quadratic program dual.

The paper demonstrates ML-based counterfactual prediction using large tax records to validate workforce training program effects via digital twins.

Implements sparse matrix exponential action to compute phylogenetic likelihoods efficiently for large state spaces.

Provides theoretical bounds and algorithmic analysis for multi-modal Laplace approximations in Bayesian sampling.

Introduces a resolution‑adaptive transformer for solar irradiance forecasting and temporal downscaling.

The paper evaluates q‑Gaussian weight and likelihood priors for Bayesian neural groundwater inversion.

The paper develops ML models using regime-conditional Granger and Student‑t HMMs to predict factor predictability decay.

It proposes a signal-processing cascade linking satellite temperature, genomic variation, and mortality risk.

Compares Gompertz‑Makeham and related mortality models using chi‑square tests on age‑specific death data.

Reviews mortality prediction models for heart failure patients using statistical, machine learning, and deep learning methods.

Provides identifiability conditions for a three‑state lexical dynamics model using aggregate corpus counts and a temporal network.

Develops a sequential Monte Carlo method for Bayesian estimation of age‑specific transmission in dynamic epidemic models.

Compares quantile regression models and calibration techniques for reliable electricity price interval forecasts.

Uses LLM to elicit behavioral priors for agent-based Bitcoin flash‑crash simulators and performs causal analysis.

Introduces scalable variational Gaussian process frameworks to correct preferential sampling for high‑resolution urban temperature mapping.

A cross‑state framework uses gradient‑boosted classifiers to detect and estimate under‑reported wildlife‑vehicle collisions from crash data.

Develops identification assumptions and statistical tests for causal effects in self‑controlled case series using case‑only data.

Defines causal effects for multiple binary treatments via odds-scaling counterfactuals and constructs efficient estimators under minimal divergence constraints.

Introduces axiomatic learning structures and transformation ranking for structure recovery from noisy data.

Develops a state‑space OU model with measurement error for longitudinal microbiome compositional data.

Constructs calibrated prediction regions for dependent censored lifetimes using conformal prediction.

Derives sharp identified intervals for survival model risk differences without target outcomes using bounded log-odds shifts.

Reconciles AIPW, TMLE, and double machine learning via shared nuisance modeling on real and synthetic data.

Characterize when stationary ergodic Markov chains satisfy a √n-CLT via finite Poisson equation solutions.

Introduces a penalised maximum likelihood method with coordinate-wise algorithm for unmixing random density mixtures in Bayes Hilbert spaces.

Introduces a robust outlier detection method for mixed continuous and ordinal data using latent Gaussian modeling.

Presents semiparametric techniques for efficient, robust estimation in the symmetric location problem.

The paper formalizes cluster-based missingness graphs and provides graphical conditions for recoverability of joint distributions and macro causal effects.

Introduces the Pivotal Information Criterion, a continuous model selection method with asymptotically pivotal penalty.

Quantifies draw non‑uniformity and implements a workable integer‑programming solution.

It provides a two-stage Bayesian method for sequential updating of spatial fragility fields using Probit‑Normal priors and GP‑ordered probit.

Implements latent space graph-fused LASSO with ADMM and Langevin dynamics for Bayesian clustering of attributed graphs.

Introduces regression discontinuity designs for causal inference in observational medical studies.

Clarifies and reformulates the no‑anticipation assumption in difference‑in‑differences causal inference.

Estimates the posterior distribution of the modal age at death using Poisson modeling and random-walk priors.

DEEM provides a debiased estimating equation for robust, efficient Mendelian randomization using many correlated SNPs.

Introduces computational equivalence and new samplers for Bayesian spatial regression models, improving inference and scalability.

Develops a CAVI algorithm for mean-field variational Bayesian scalar-on-function regression and a global test for functional effects.

Blends AI and physics ensemble forecasts to boost UK weather forecast skill via CRPS analysis.

Derives sparsity‑based co‑sufficient weights for weighted Holm and BH multiple‑testing procedures.

Develops a semiparametric estimator for missing-not-at-random data integration using instrumental variables and cross‑fitted orthogonal equations.

Estimates observation-specific bias and variance in latent-variable regression using Gaussian processes.

Derives minimax MSE bounds for estimating shared treatment effects under limited overlap using functional models.

Constructs spatial sampling designs with equal inclusion probabilities that yield different prediction risk rankings.

Introduces a hierarchically weighted conformal prediction engine for auditable per‑cell interval predictions in battery SOH monitoring.

Provides positive‑semidefinite covariance estimators for sample L‑moments with faster rank‑prefix algorithms.

Provides a unified uncertainty quantification framework for topological treatment effects with confidence bands and prediction intervals.

RW-DPO introduces reliability-weighted DPO with a minimax penalty to robustly learn from noisy preference data.

The study uses ICP-MS mineral profiles and machine learning to discriminate Beaujolais wine crus.

Provides a unified lemma and eigenvalue-based proof for identifying shocks via heteroskedasticity patterns.

Constructs admissible e-value contracts that incentivize truthful parameter reporting under nuisance uncertainty.

Proposes a three-stage transfer learning estimator for PDE solution operators using instrumental variables and sparse calibration.

Introduces Source Projection Transfer to calibrate survival predictions via projected source scores onto target models.

Develop balancing estimators for continuous treatments via continuity and basis function criteria, with variance efficiency guarantees.

Proposes a GLM for covariate effects on the quantile correlation coefficient of bivariate outcomes.

Proposes marginal effect measures to interpret stochastic actor-oriented network model parameters.

Analyzes information boundaries in loss reserving using paid triangles to quantify reserve uncertainty.

It provides a scalable approximate method to combine subposterior distributions for distributed Bayesian inference.

Demonstrates quantum amplitude amplification to increase probability of rare simulation outcomes.

Provides a spectral hypergraph cycle enumeration method for efficient planted k‑XOR inference.

Combines marginal and conditional evidence to test global null of zero means in correlated Gaussian data.

Learns a smeared density ratio to define signal regions without strong signal or background assumptions.

Proposes fuzzy entropic triple k-means for clustering three-way data arrays via entropy-regularized fuzzy membership and block centroids.

Proposes a discordance-based sensitivity model for DiD, linking parallel pretrends to alternative imputations via expected ATT distance.

Proposes budgeted BONuS-GA to control finite-sample FDR in distributed network multiple testing with limited communication.

They compare and improve GEDI waveform simulation against ALS data to reduce bias in forest height estimates.

Estimates correlated gravitational‑wave detector noise spectral density via a matrix‑gamma process prior and MCMC.

Derives closed-form vector-sum rules for perceived vertical orientation from line angles.

Uses large language models with in-context learning to impute missing responses in public opinion surveys.

Develops identifiability analysis methods for partially observed stochastic processes using trajectories and density data.

Provides an R package (INLAvaan) for fast approximate Bayesian structural equation modelling using Integrated Nested Laplace Approximation.

Introduces a bias‑free rotation method for latent representations with oracle‑level inference guarantees.

Integrates design-based calibration with Fay‑Herriot small area estimation using incomplete auxiliary data.

The paper proposes a scalable framework for causal mediation analysis in large DAGs with multiple treatments and mediators.

Integrates an AE ontology into a zero‑inflated negative binomial model for ADR detection with a data‑thinning validation technique.

Reviews rerandomization theory, simulates finite‑sample performance, and provides a practitioner guide.

Provides a concise mathematical introduction to Monte Carlo methods and their algorithmic foundations.

We develop semiparametric efficient two-stage estimators for causal excursion effects with robustness to nuisance model misspecification.

Constructs coordinate‑free latent spaces from ordinal data using projective geometry to achieve ratio‑scale measurement and resolve identifiability.

PCMCI+ extends PCMCI to discover contemporaneous causal links in autocorrelated nonlinear time series via optimized conditioning sets.

Proves asymptotic Gaussianity for heterogeneous spatial autoregressions on sparse graphons, enabling network‑free inference.

The paper evaluates hybrid GLM, GAM, and deep regression models for spatio‑temporal historical map data.

Derives activation boundary for weak factors under exchangeable shrinkage prior in sparse Bayesian factor models.

Constructs uncertainty graphs for distribution‑free predictive ordering with error‑rate control.

Proposes an oracle weighted neural estimator with minimax optimal rates for high-dimensional nonparametric regression under multisource covariate shift.

Posterior-landscape extracts and characterizes 2D density features while propagating Bayesian uncertainty across posterior samples.

Introduces a layered grammar of graphics framework to compose spatial uncertainty visualisations in ggplot2.

CARED uses randomized expert review to calibrate AI labels and control cumulative error under distribution shift.

Introduce hybrid hierarchical matrices enabling fast kernel matrix-vector products for Gaussian processes and kernel ridge regression.

Proposes a hierarchical sparse group recovery method for panel data to jointly identify latent groups and sparse parameters.

Proposes ensemble null tests to compare multivariate distributions for detecting systematic errors beyond scalar statistics.

It derives decomposable graphical models from irreversible event accumulation on latent rooted trees, linking cliques to tree edges.

Constructs joint distributions from marginals and a graph using optimal transport to enforce conditional independence.

Develops policy learning and off‑policy estimation methods for reinforcement learning with runtime confounding.

Develops adaptive sequential Bayesian spatio‑temporal surveillance using posterior thresholds and inverse‑probability weighting.

Sequince provides a modular Python toolkit for building sequential Monte Carlo algorithms with reusable components.

The paper proposes a Bayesian predictive synthesis using softmax weight models and bias‑adjusted densities, with Pólya‑gamma Metropolis‑Hastings for posterior computation.

Develops a reference-based multiple imputation framework for longitudinal bounded discrete outcomes using Kumaraswamy margins and Bayesian latent variable augmentation.

The paper compares confidence-based and random selection for video gait data, assessing measurement error and selection shift effects.

Provides a computationally efficient infinitesimal-jackknife variance matching method to calibrate generalized posterior learning rates for accurate credible intervals.

Introduces a variational Bayesian variable selection method with adaptive sharing across populations and responses, scalable to summary data.

Develops a multi‑scale stochastic framework with likelihood‑free Bayesian calibration and stochastic finite elements to link interfacial variability to panel response.

The paper proposes a Bayesian dynamic programming policy for MLB challenge decisions using pitch location uncertainty.

Proposes nonlinear sufficient dimension reduction to improve causal effect estimation with continuous treatments.

Proposes a self‑normalised change‑point test that stays pivotal under non‑stationary noise.

Proves optimal expectation bounds for tensor fiber norms using matrix norm comparisons.

Develops weighted score matching for nonparametric density estimation and interaction graph recovery on bounded domains.

Introduces nonparametric independence tests for general spatial point processes using characteristic function differences without random shifts.

TwinPoSI generates statistical twins to enable accurate post‑selection confidence intervals for population‑based regression coefficients.

Introduces Bregman projection-based partition design to accelerate finite Markov chain mixing.

We introduce a sparse variational inference method for estimating a latent VAR network from multivariate count time series.

Jointly optimize interim timing and stopping boundaries to minimize expected sample size under error rate constraints in group sequential trials.

The paper evaluates a new wins above replacement estimator against existing metrics using historical baseball data.

Develops a statistical framework to jointly model screening and randomisation and determine adaptive stopping times for clinical trials.

The paper proposes a taxonomy of reliability and instability mechanisms for LLM-assisted evidence synthesis tasks.

Evaluates calibration of tabular foundation models versus logistic regression in simulated and clinical risk prediction.

Modular Bayesian foundation models use uncertainty-guided routing of compact SBNN encoders for efficient time‑series forecasting and OOD detection.

Develops ordered-gap estimator with direction-specific concentration radii for exact recovery of latent groups from multiview ordinal ranks.

Optimizes a convex mixed-integer program to identify invariant predictors across multiple environments.

The paper extends Bayesian discrepancy modeling with a stick‑breaking EPR method for multinomial data to efficiently analyze high‑dimensional land‑water transitions.

Uses generative AI to compare flood loss reduction from hard versus soft urban infrastructure.

Introduces Proximal R2 priors that encode structural variance constraints via proximal relaxations, enabling differentiable posterior inference.

Extends principal‑agent hypothesis testing to sequential settings, characterizing null‑deterring contracts with e‑processes and optimal stopping via Snell envelope.

DECIDE quantifies individual decision impacts in multiverse analyses without model prediction.

Introduces Metropolized backward sampling to improve particle Gibbs sampling in state‑space models.

Integrates multi-source partial graph data via integer programming to reconstruct a global undirected graphical model.

Estimates a mixed geographically weighted spatial autoregressive model for budget hotels using satellite-derived environmental and transport indicators.

Evaluates power‑normal distribution for flexible approximation of unimodal positive distributions and provides an accompanying R package.

The paper introduces an optimal approximation error framework to correct Bayesian posterior bias via adjusted likelihoods.

The paper introduces a compartment‑driven Gaussian process model for Bayesian uncertainty quantification of pharmacokinetic exposure metrics.

Develops a robust quadratic score test for mean functions in functional and longitudinal data via convex loss and spline approximation.

Derives group sequential boundaries for time‑to‑event endpoints using joint combination statistic distribution and numerical integration.

Evaluates strip trial designs, showing replication drives power and spatial models enhance detection.

Introduces dual optimization for AIS trajectory simplification, jointly optimizing compression and length error via dynamic programming.

Introduces a simple sensitivity analysis to assess causal effect robustness from observational data using point estimates and confidence intervals.

FACTS-MAS provides a multi-agent framework to combine heterogeneous forecasting evidence via weighting and calibration.

Introduces weight-changing hazard estimator for IPCW survival estimation with improved bias‑variance tradeoff.

L-PASSO is an adaptive log‑penalized regression method that achieves robust sparse variable selection with provable oracle properties.

The study shows that few hundred engine games yield unstable opening rankings due to sampling error and draws.

Provides a tree-based uplift model with certified FWER control for treatment effect claims.

The paper presents fastC-GC and fastC-GC* algorithms for efficient time‑series graph estimation using ridge‑residual correlation and Fourier‑based shift enumeration.

Derives uniform 2‑Wasserstein bounds for noisy‑gradient Euler and randomized midpoint Langevin samplers.

The paper proposes a hierarchical Bayesian mixture model to combine incomplete surveys and estimate galaxy flux distributions.

Proposes a simulation‑based two‑stage stochastic optimization procedure for model criticism without pre‑specified test statistics.

The paper creates a unified, deduplicated aviation safety incident dataset from 128 sources.

Develops an influence-function estimator for marginal mean models with flexible link functions and data-driven sensitivity parameter selection.

It compares MLE and Bayesian methods for global parameter identification and uncertainty quantification of a reduced-order ADM1 model.

Introduces folded concave penalties to improve variable selection and prediction in cointegrated regression.

Anchors e-closure to transparently enhance FDR multiple testing procedures while preserving FDR control.

Flow matching learns a transport map to generate approximate posterior samples for Bayesian inverse problems with negligible online cost.

Introduces a global synchronization algorithm for jointly aligning multiple low-rank matrices with missing data.

We derive debiased estimators for optimal value in RL via softmax‑approximated Bellman fixed points.

Derives an approximate MSE criterion to select dictionary and penalty for instrumental variables LASSO.

Develops fast, order‑invariant equation‑by‑equation Bayesian VAR estimation using factor‑structured reduced‑form errors for conditional forecasting and scenario analysis.

Introduces Bayesian inference for Dirichlet process centering measure and precision via data‑generating model.

Derives asymptotic theory and corrected standard errors for granular instrumental variables in large panels.

The paper critiques statistical errors in a forensic analysis of nurse roster charts.

The paper introduces a panel of large language models treated as raters to measure political positions in data‑sparse regions.

DEAL provides bias-aware, cross-fitted debiased Lasso intervals that are shorter and more accurate in high-dimensional regression.

Express converts non-causal attention to causal with matching guarantees, reducing memory and compute costs.

Derive efficient influence function and doubly robust estimator for functional treatment effects in DiD.

Introduces a dynamic latent community framework for VAR-type models using co-blockmodel embedding and spectral co-clustering.

Introduces convolutional MMD for simulation‑based inference with noisy data, enabling robust, likelihood‑free estimation.

Introduces a stochastic T-process model for real-time win probability and latent player contribution via STATS X.

Introduces an estimand framework to assess transportability of indirect treatment comparisons in health technology assessment.

Provides a contribution-adjusted Bayesian meta-analysis framework for subgroup-specific effects and interactions.

Introduces a Bayesian graphical model linking latent positions to persistent homology bar events for inference.

Provides closed-form bounds and flexible estimators for the average derivative effect under a sensitivity model for continuous exposures.

Proposes a Grassmannian gradient descent algorithm for matrix completion with dependent missingness and provides debiased uncertainty quantification.

Proposes a doubly robust estimator with stabilized weights for binary outcomes in micro-randomized trials.

WISER uses epidemic change-point modeling to efficiently segment watermarked text regions.

Introduces a hierarchical Gaussian process prior for covariate-driven Poisson processes with adaptive posterior concentration and MCMC inference.

Introduces the Same Weighting Across Different Analyses principle for coherent subgroup and interaction meta‑analysis estimates.

Derives corrected spherical‑Dirichlet density and analyzes its moments, mode, and properties.

Estimates auxiliary outcomes to correct regression standard errors for spatial correlation.

Introduces LSVI, a gradient‑free variational inference algorithm based on least‑squares updates.

Introduces full conformal e-values for distribution‑free novelty detection with provable FDR control.

Introduces a differentially private weighting estimator for causal effect inference with covariate balance.

Conditionally calibrates e-values to boost e-BH power while preserving FDR control.

Introduces robustified modifications of classification metrics to handle class imbalance.

The paper develops an integrative multiview structural equation model for microbiome-metabolome-disease pathways with theoretical guarantees.

Estimates optimal policy for dynamic strategy selection via regularized iterated regression in a reinforcement learning framework.

The paper develops a kernel-based regularized least squares approach for Volterra series modeling and forecasting of time series.

Derives minimax confidence interval rates for monotone multivariate nonparametric regression under Gaussian errors.