This paper develops asymptotic theory for quantile estimation via stochastic gradient descent (SGD) with a constant learning rate. The quantile loss function is neither smooth nor strongly convex. Beyond conventional perspectives and techniques, we view quantile SGD iteration as an irreducible, periodic, and positive recurrent Markov chain, which cyclically converges to its unique stationary distribution regardless of the arbitrarily fixed initialization. To derive the exact form of the stationary distribution, we analyze the structure of its characteristic function by exploiting the stationary equation. We also derive tight bounds for its moment generating function (MGF) and tail probabilities. Synthesizing the aforementioned approaches, we prove that the centered and standardized stationary distribution converges to a Gaussian distribution as the learning rate $\eta\rightarrow0$. This finding provides the first central limit theorem (CLT)-type theoretical guarantees for the quantile SGD estimator with constant learning rates. We further propose a recursive algorithm to construct confidence intervals of the estimators with statistical guarantees. Numerical studies demonstrate the effective finite-sample performance of the online estimator and inference procedure. The theoretical tools developed in this study are of independent interest for investigating general SGD algorithms formulated as Markov chains, particularly in non-strongly convex and non-smooth settings.

Multiple testing adjustments, such as the Benjamini & Hochberg (1995) step-up procedure for controlling the false discovery rate (FDR), are typically applied to families of tests that control significance level in the classical sense: for each individual test, the probability of false rejection is no greater than the nominal level. In this paper, we consider tests that satisfy only a weaker notion of significance level control, in which the probability of false rejection need only be controlled on average over the hypotheses. We find that the Benjamini & Hochberg (1995) step-up procedure still controls FDR in the asymptotic regime with many weakly dependent p-values and an increasing number of rejections, and that certain adjustments for dependent p-values such as the Benjamini & Yekutieli (2001) procedure continue to yield FDR control in finite samples. Our results open the door to FDR controlling procedures in nonparametric and high dimensional settings where weakening the notion of inference may allow for power improvements.

This paper studies the hypothesis testing problem of deciding whether $m \geq 2$ complete weighted graphs with Gaussian edge weights are mutually correlated after unknown relabelings of their vertices. Under the null model all edge weights are independent standard Gaussians, whereas under the planted model the graphs share a latent vertex alignment and each pair of corresponding edge weights has correlation $\rho$. For fixed $m$, we identify the sharp information-theoretic threshold for detection. Above the threshold, a generalized likelihood-ratio test achieves strong detection, whereas even weak detection is impossible below the threshold. The result extends the two-graph detection threshold of Wu, Xu, and Yu to any fixed number of graphs, exhibits a side-information regime in which two graphs alone are insufficient but multiple graphs enable detection, and, together with the recovery threshold of Vassaux and Massoulié, shows that this Gaussian multi-graph model has no detection--recovery gap.

Online conformal prediction must balance fast adaptation to distribution shift against stable coverage: feedback-driven methods react quickly but become volatile, while strongly discounted Bayesian methods lag and inflate intervals at tight coverage. We introduce \textbf{State-Adaptive Bayesian Conformal Prediction (SA-BCP)}, which forms the predictive quantile as a gated convex combination of long-term temporal inertia and local spatial evidence from a kernel density estimate, controlled by a single interpretable evidence threshold $K$. We establish three results: (i) asymptotic marginal validity of the resulting intervals; (ii) a closed-form expression for the MSE-optimal threshold, $K^*_{\mathrm{MSE}}=\alpha(1-\alpha)/M^{\mathcal{T}}$, trading the coverage-indicator (Bernoulli) variance against the temporal structural bias $M^{\mathcal{T}}$; and (iii) a rolling-origin procedure for selecting $K$ online -- consistent under stationarity, with $O(\sqrt{T\log N})$ regret against the best fixed $K$ and, for a segmented variant, a sublinear dynamic-regret bound under bounded drift. Across four financial-volatility and weather datasets, three target coverage levels, and eight baselines (including the strongest recent conditional-quantile methods, SPCI and KOWCPI), SA-BCP attains at-or-above-nominal coverage in most settings while producing substantially sharper intervals -- up to roughly $3\times$ lower Winkler score than discounted Bayesian CP at the tightest coverage -- and a coverage-matched audit confirms these efficiency gains are not an artifact of under-coverage. We disclose one principal limitation: a volatility-specialized conformal-GARCH competitor remains more efficient on its home volatility-base series, though it does not transfer across domains.

High-dimensional, underdetermined and highly correlated systems are common in data science practice, especially when analyzing physical measurements. In such settings, feature selection poses a fundamental challenge because multiple distinct sparse subsets may explain the response equally well. Their identification is crucial not only for predictive modeling but also for generating domain-specific insights into the underlying mechanisms. Yet, conventional methods typically isolate a single solution, obscuring the full spectrum of plausible explanations. This work introduces GEMSS (Gaussian Ensemble for Multiple Sparse Solutions), a variational algorithm designed to simultaneously discover multiple, diverse sparse feature combinations. The method employs a structured spike-and-slab prior for sparsity, a mixture of Gaussians to approximate the intractable multimodal posterior, and a Jaccard-based penalty to further control solution diversity. A single objective function is optimized via stochastic gradient descent. The method is tested on 128 comprehensive experiments by a novel benchmarking framework designed to generate artificial problems with multiple sparse solutions of equal predictive properties. This allows us to measure the retrieval of ground truth features rather than only evaluating predictive performance -- characteristics more fitting to our practical needs. A comparative analysis shows that GEMSS consistently outperforms five prominent feature selection methods adapted through the ALFESE framework. Finally, we demonstrate practical usability through 3 challenging real-world datasets from metabolomics and physical chemistry: GEMSS successfully isolates multiple distinct yet quality solutions. GEMSS is available as a PyPI package 'gemss'. The corresponding repository this http URL includes the full codebase and a free, no-code application GEMSS Explorer.

Online joint estimation of a dynamical model's unknown parameters and states with uncertainty quantification is crucial in many applications. For example, digital twins dynamically update their knowledge of model parameters and states to support prediction and decision-making. Reliability and computational speed are vital for DTs. Online parameter-state estimation ensures computational efficiency, while uncertainty quantification is essential for making reliable predictions and decisions. In parameter-state estimation, the joint distribution of the state and model parameters conditioned on the data, termed the joint posterior, provides accurate uncertainty quantification. Because the joint posterior is generally intractable to compute, this paper presents an online variational inference framework to compute its approximation at each time step. The approximation is factorized into a marginal distribution over the model parameters and a state distribution conditioned on the parameters. This factorization enables recursive updates through a two-stage procedure: first, the parameter posterior is approximated via variational inference; second, the state distribution conditioned on the parameters is computed using Gaussian filtering based on the approximate parameter posterior. The algorithmic design is supported by a theorem establishing upper bounds on the joint posterior approximation error. Numerical experiments demonstrate that the proposed method (i) accurately infers both unobserved states and unknown parameters of dynamical and observation models; (ii) remains robust under noisy, partial observations and model discrepancies in a chaotic Lorenz'96 system; and (iii) scales effectively to a high-dimensional state-space system arising from the spatial discretization of a convection-diffusion equation. outperforming the joint ensemble Kalman filter in this setting.

Nonparanormal models describe the joint distribution of multivariate responses via latent Gaussian, and thus parametric, copulae while allowing flexible nonparametric marginals. Some aspects of such distributions, for example conditional independence, are formulated parametrically. Other features, such as marginal distributions, can be formulated non- or semiparametrically. Such models are attractive when multivariate normality is questionable but interpretability paramount. Most estimation procedures perform two steps, first estimating the nonparametric part. The copula parameters come second, treating the marginal estimates as known. This is sufficient for some applications. For other applications, e.g. when a semiparametric margin features parameters of interest or when standard errors are important, a simultaneous estimation of all parameters might be more advantageous. We present suitable parameterisations of nonparanormal models, possibly including semiparametric effects, and define four novel nonparanormal log-likelihood functions. In general, the corresponding one-step optimisation problems are shown to be non-convex. In some cases, however, biconvex problems emerge. Several convex approximations are discussed. From a low-level computational point of view, the core contribution is the score function for multivariate normal log-probabilities computed via Genz procedure. As a demonstration for the versatility of the theoretical and computational framework, we present a series of nonparanormal models for transformation discriminant analysis when some biomarkers are subject to limit-of-detection problems. Possible empirical gains of full maximum likelihood estimation compared to two-step approaches are illustrated in a simulation study targeting semiparametric efficient polychoric correlation analysis where a theoretical benchmark is available.

The Poisson compound decision problem is a long-standing problem is statistics, for which empirical Bayes methods are commonly used to estimate Poisson means in static or batch settings. We consider this problem in a streaming, or online, framework. Building on a quasi-Bayesian approach based on Newton's algorithm, we develop a sequential estimate that is easy to evaluate, computationally efficient, and has constant per-observation cost as the data accrue. We establish frequentist guarantees for the proposed estimate, including consistency and asymptotic optimality, with optimality understood as vanishing excess Bayes risk, or regret. Empirical performance is assessed through simulation studies and comparisons with benchmark procedures.

Two-sample testing is a fundamental tool for detecting distributional differences across scientific domains, but classical tests (including kernel-based tests) can be ineffective on high-dimensional structured data such as images. Recent deep two-sample tests improve sensitivity in these settings by learning informative representations, yet they provide limited insight into which data features drive rejection of the null hypothesis $H_0$. To address this issue, we propose a counterfactual explanation framework for deep two-sample testing that generates sample-level edits moving observations from a source group toward a target group while explicitly reducing the discrepancy measured by the test. Our method combines a diffusion autoencoder with a pretrained deep two-sample test model and optimizes a maximum mean discrepancy (MMD) objective in the test model's representation space to produce plausible counterfactuals. We quantify distribution-level effects through changes in the test statistic and the resulting two-sample p-values. We evaluate the method on synthetic 2D shape datasets and two MRI cohorts. Across both settings, the counterfactual transformations consistently increase p-values relative to the original samples, indicating that the edited source set becomes statistically closer to the target distribution under the test. We measure minimality using LPIPS to ensure the counterfactuals remain close to the original samples. The resulting edits provide interpretable evidence of the features associated with the detected group differences. On MRI, the localized changes are consistent with known anatomical differences between cohorts.

Causal mediation analysis decomposes the total treatment effect into a portion operating through a hypothesized mediator and a residual direct portion. Identification of natural direct and indirect effects typically rests on the mediator stage of sequential ignorability, which cannot be empirically verified and requires explicit sensitivity analysis. We formulate the \emph{bridge score}, a mediator-stage balancing score, as a low-dimensional vector formed from the two treatment-specific mediator densities at a common mediator value, and show that it balances baseline covariates for the mediator stage relevant to natural effect identification. Conditional on the bridge score, we derive a sharp pointwise variance envelope on the unidentified mediator-outcome confounding function in terms of latent outcome relevance and residual selection. To make the bound operational for sensitivity analysis, we further introduce a residual budget calibration approach based on local residual outcome variation and record a complementary range bound for support-based restrictions. Finally, we show how the pointwise bound can be operationalized for inference through a scalar functional reduction and a Bayesian g-computation algorithm that combines observed-data posterior uncertainty with user-specified sensitivity uncertainty, rather than treating the unidentified sensitivity corrections as learned from the likelihood.

Evaluating the causal health effects of multivariate, continuous exposures, such as air pollution mixtures, is a critical public health challenge. A primary obstacle is the frequent violation of the positivity assumption, which renders the effects of standard deterministic interventions unidentified or heavily reliant on unreliable model extrapolation. In this paper, we develop a novel causal inference framework to address this challenge. We extend exponential tilting to multivariate exposures and address the critical question of how to compare different intervention directions fairly. This establishes a systematic framework for defining and evaluating various policy-relevant causal estimands, allowing researchers to address diverse scientific questions. We develop numerous methodological advancements, including efficient one-step estimation strategies, a Riemannian BFGS algorithm to solve a constrained manifold optimization problem, semiparametric efficiency bounds for causal estimands, minimax rates for estimators, and establishing asymptotic normality. We demonstrate our framework's utility by applying it to a nationwide environmental health dataset to identify the optimal strategy for reducing adverse health outcomes associated with a PM$_{2.5}$ chemical mixture.

In this paper, we demonstrate how to enhance the validity of causal inference with unstructured high-dimensional treatments like texts, by leveraging the power of generative Artificial Intelligence (GenAI). Specifically, we propose to use a deep generative model such as large language models (LLMs) to efficiently generate treatments and use their internal representation for subsequent causal effect estimation. We show that the knowledge of this true internal representation helps disentangle the treatment features of interest, such as specific sentiments and certain topics, from other possibly unknown confounding features. Unlike existing methods, the proposed GenAI-Powered Inference (GPI) methodology eliminates the need to learn causal representation from the data, and hence produces more accurate and efficient estimates. We formally establish the conditions required for the nonparametric identification of the average treatment effect, propose an estimation strategy that avoids the violation of the overlap assumption, and derive the asymptotic properties of the proposed estimator through the application of double machine learning. Finally, using an instrumental variables approach, we extend the proposed GPI methodology to the settings in which the treatment feature is based on human perception. The GPI is also applicable to text reuse where an LLM is used to regenerate existing texts. We conduct simulation and empirical studies, using the generated text data from an open-source LLM, Llama 3, to illustrate the advantages of our estimator over state-of-the-art causal representation learning algorithms.

This paper studies a two-stage model of experimentation, where the researcher first samples representative experimental participants from an eligible pool, then assigns each sampled unit to treatment or control, using matched $k$-tuples randomization at both stages. To implement such designs, we develop a fast new algorithm for matching units into $k$-tuples for any $k \ge 2$ and any dimension of covariates. By surveying 200 recent experimental working papers, we estimate that our algorithm newly enables multivariate fine stratification with provable match quality guarantees for about 44\% of experiments in economics. We show that finely stratified sampling and assignment both nonparametrically reduce the variance of treatment effect estimation, with the gains from stratified sampling increasing in the size of the eligible pool and how well covariates predict treatment effect heterogeneity. We develop new inference methods that fully exploit the efficiency gains from both design stages, allowing researchers to report smaller standard errors if they designed a representative experiment. An application to nine published experiments quantifies the efficiency gains.

Importance sampling is a Monte Carlo method which designs estimators of expectations under a target distribution using weighted samples from a proposal distribution. When the target distribution is complex, such as multimodal distributions in highdimensional spaces, the efficiency of importance sampling critically depends on the choice of the proposal distribution. In this paper, we propose a novel adaptive scheme for the construction of efficient proposal distributions. Our algorithm promotes efficient exploration of the target distribution by combining global sampling mechanisms with a delayed weighting procedure. The proposed weighting mechanism plays a key role by enabling rapid resampling in regions where the proposal distribution is poorly adapted to the target. Our sampling algorithm is shown to be geometrically convergent under mild assumptions and is illustrated through various numerical experiments.

The problem of optimising functions with intractable gradients frequently arises in machine learning and statistics, ranging from maximum marginal likelihood estimation procedures to fine-tuning of generative models. Stochastic approximation methods for this class of problems typically require inner sampling loops to obtain (biased) stochastic gradient estimates, which rapidly becomes computationally expensive. In this work, we develop sequential Monte Carlo (SMC) samplers for optimisation of functions with intractable gradients. Our approach replaces expensive inner sampling methods with efficient SMC approximations, which can result in significant computational gains. We establish convergence results for the basic recursions defined by our methodology which SMC samplers approximate. We demonstrate the effectiveness of our approach on the reward-tuning of energy-based models within various settings.

How can we learn the laws underlying the dynamics of stochastic systems when their trajectories are sampled sparsely in time? Existing methods either require temporally resolved high-frequency observations, or rely on geometric arguments that apply only to conservative systems, limiting the range of dynamics they can recover. Here, we present a new framework that reconciles these two perspectives by reformulating inference as a stochastic control problem. Our method uses geometry-driven path augmentation, guided by the geometry in the system's invariant density to reconstruct likely trajectories and infer the underlying dynamics without assuming specific parametric models. Applied to overdamped Langevin systems, our approach accurately recovers stochastic dynamics even from extremely undersampled data, outperforming existing methods in synthetic benchmarks. This work demonstrates the effectiveness of incorporating geometric inductive biases into stochastic system identification methods.

Plug-and-Play (PnP) algorithms are a class of iterative algorithms that address image inverse problems by combining a physical model and a deep neural network for regularization. Even if they produce impressive image restoration results, these algorithms rely on a non-standard use of a denoiser on images that are less and less noisy along the iterations, which contrasts with recent algorithms based on Diffusion Models (DM), where the denoiser is applied only on re-noised images. We propose a new PnP framework, called Stochastic deNOising REgularization (SNORE), which applies the denoiser only on images with noise of the adequate level. It is based on an explicit stochastic regularization, which leads to a stochastic gradient descent algorithm to solve ill-posed inverse problems. A convergence analysis of this algorithm and its annealing extension is provided. Experimentally, we prove that SNORE is competitive with respect to state-of-the-art methods on deblurring and inpainting tasks, both quantitatively and qualitatively.

We investigate the convergence properties of popular data-augmentation samplers for Baye\-sian probit regression. Leveraging recent results on Gibbs samplers for log-concave targets, we provide simple and explicit non-asymptotic bounds on the associated mixing times (in Kullback-Leibler divergence). The bounds depend explicitly on the design matrix and the prior precision, while they hold uniformly over the vector of responses. We specialize the results for different regimes of statistical interest, when both the number of data points $n$ and parameters $p$ are large: in particular we identify scenarios where the mixing times remain bounded as $n,p\to\infty$, and ones where they do not. The results are shown to be tight (in the worst case with respect to the responses) and provide guidance on choices of prior distributions that provably lead to fast mixing. An empirical analysis based on coupling techniques suggests that the bounds are effective in predicting practically observed behaviours.

Modeling unknown latent functions from finite, irregularly sampled measurements is a recurring challenge across science and engineering. Neural processes (NPs), a family of probabilistic functional models, are promising solutions -- especially when endowed with domain-specific symmetries like translation equivariance, which improve sample efficiency and generalization. Yet existing translation-equivariant NPs face two limitations: (i) they stack generic components with non-linearities, obscuring the induced function class and limiting interpretability; and (ii) convolutional designs rely on kernels with local receptive fields and require dense uniform input grids, while attention-based methods avoid these issues but scale quadratically with the number of observations. We address both with two contributions. First, using the Volterra expansion, we characterize continuous translation-equivariant operators as sums of higher-order convolutions, yielding analytical transparency while admitting efficient approximation by first-order convolutions. Second, we introduce set Fourier convolutions (SFConvs), a frequency-domain parameterization that operates directly on irregularly sampled points, achieves approximately global receptive fields, and scales linearly in the number of observations. Building on these ideas, we propose two conditional NPs (CNPs): SFConvCNPs, which stack SFConv blocks with non-linearities, and SFVConvCNPs, which integrate the Volterra formulation. Experiments on synthetic and real-world datasets demonstrate our methods' efficacy against state-of-the-art baselines.

Many problems in computational science and engineering become one-to-many after coarse graining, partial observation, or inverse reconstruction: a resolved state may not determine a unique subgrid forcing, a structural descriptor may not determine a unique effective response, and a low-resolution observation may correspond to many plausible high-resolution fields. In such settings, deterministic surrogates may learn a well-defined mathematical object while still missing application-relevant uncertainty. This tutorial develops a self-contained module centered on the conditional-mean barrier: the point at which a squared-loss predictor has reached the conditional mean and the remaining error is irreducible aleatoric variance. We give two diagnostics for locating this barrier, residual-feature orthogonality and the coefficient of determination against its explained-variance ceiling, and prove that adding latent randomness to a squared-loss predictor collapses it back to the conditional mean. Crossing the barrier therefore requires a loss that scores distributions rather than point predictions. We briefly organize common distributional objectives, including negative log-likelihood, moment and observable matching, variational objectives, adversarial divergences, and score matching, by the feature of the conditional law each targets. The emphasis is the boundary itself and a finite-data procedure for recognizing it, rather than a survey of methods beyond it. CPU-based demonstrations on a two-branch law and a two-scale Lorenz-96 closure problem show how the diagnostics distinguish deterministic underfitting from residual distributional variability.

Mirror Descent (MD) is a scalable first-order method widely used in large-scale optimization, with applications in image processing, policy optimization, and neural network training. This paper generalizes MD to optimization on Riemannian manifolds. In particular, we develop a Riemannian Mirror Descent (RMD) framework via reparameterization and further propose a stochastic variant of RMD. We also establish non-asymptotic convergence guarantees for both RMD and stochastic RMD. As an application to the Stiefel manifold, our RMD framework reduces to the Curvilinear Gradient Descent (CGD) method proposed in [26]. Moreover, when specializing the stochastic RMD framework to the Stiefel setting, we obtain a stochastic extension of CGD, which effectively addresses large-scale manifold optimization problems.

Evaluating and interpreting latent representations, such as variational autoencoders (VAEs), remains a significant challenge for diverse data types, especially when ground-truth generative factors are unknown. To address this, we unify several state-of-the-art disentangled VAE approaches for latent space disentanglement into one framework -- bfVAE. To assess the effectiveness of a disentangled VAE model and enhance latent space interpretability, we propose Feature Variance Heterogeneity via Latent Traversal (FVH-LT) and Dirty Block Sparse Regression in Latent Space (DBSR-LS). To ensure robust interpretability of learned latent space, we develop a greedy alignment strategy (GAS) that mitigates label switching and aligns latent dimensions across runs to set the foundation of result aggregation. We also introduce a convenient scalar latent space separation index (LSSI) based on the GAS-aligned outputs of FVH-LT and DBSR-LS to summarize the overall latent structural separation without knowledge of the ground-truth generative factors. We compare bfVAE to five VAE models and validate the effectiveness FVH-LT, DBSR-LS, and LSSI in on seven tabular and image datasets. Under our examined experimental settings, bfVAE provides a more flexible disentanglement framework achieves more favorable overall trade-off between disentanglement and reconstruction than the benchmark VAE models; FVH-LT and DBSR-LS reliably uncover semantically meaningful and domain-relevant latent structures and generally yield consistent results; and LSSI makes an effective quantitative summary of latent structural separation.

The RemOve-And-Retrain (ROAR) benchmark is widely used to evaluate feature attribution methods, yet its validity remains underexplored from an information-theoretic perspective. We show that model- and data-agnostic post-processing of attribution maps (transformations that, by the data processing inequality, \emph{cannot} add information about the decision function) can often improve ROAR scores. This means that an improved ROAR ranking is not, by itself, evidence that an attribution map carries more information about the model. We trace this failure mode to a bias toward spatially blurry masks. Experiments on CIFAR-10, SVHN, and CUB-200 show a consistent association between blurriness and ROAR performance, a pattern that also appears in the ROAD variant. We provide guidelines for more cautious removal-based benchmarking, with implications for validating mechanistic understanding of neural network internals.

An adaptive bandwidth selection procedure for the mixture kernel in the maximum mean discrepancy (MMD) for fitting generative moment matching networks (GMMNs) is introduced, and improved learning of copula random number generators is demonstrated. Based on the relative error of the training loss, the number of kernels is increased during training; additionally, the relative error of the validation loss is used as an early stopping criterion. While training time remains similar, adaptively training GMMNs (AGMMNs) significantly increases training performance, which is shown based on validation MMD trajectories, samples and validation MMD values. Superiority of AGMMNs over GMMNs and parametric copula models is also demonstrated in terms of three applications. First, convergence rates of estimators based on quasi-random versus pseudo-random samples from copulas are investigated in dimensions as large as 100 for the first time. Second, replicated validation MMDs, as well as Monte Carlo and quasi-Monte Carlo applications demonstrate the improved training of AGMMNs for a copula model implied by the 50 constituents of the S&P 500 index after deGARCHing. Last, both the latter dataset and 50 constituents of the FTSE 100 are used to demonstrate that the improved training of AGMMNs indeed translates to an improved model prediction.

The architecture of a neural network and the choice of its activation function are both fundamental to its performance. Equally important is ensuring that these two elements are well matched, as their alignment is key to effective representation and learning. In this paper, we introduce the Fourier Multi-Component and Multi-Layer Neural Network (FMMNN), a model that combines sine-type activations with the multi-component and multi-layer structure of MMNNs. In an FMMNN, each component is represented as a trainable linear combination of fixed random sine-type basis functions, while multi-layer composition generates more complex and adaptive high-frequency features. We establish that FMMNNs retain exponential expressive power for function approximation even under a low-rank architectural structure. We also analyze the optimization landscape of FMMNNs and find it to be substantially more favorable than that of standard fully connected neural networks, especially for high-frequency targets. In addition, we propose a scaled random initialization method for the first-layer weights in FMMNNs, which accelerates training and improves final performance when sufficient samples are available. Extensive numerical experiments support our theoretical insights, showing that FMMNNs achieve strong accuracy and favorable convergence behavior on oscillatory function-approximation benchmarks.

Population-adjusted indirect comparisons (PAICs) are widely used to synthesize evidence when randomized controlled trials enroll different patient populations and head-to-head comparisons are unavailable. Although PAICs adjust for observed population differences across trials, adjustment alone does not ensure transportability of estimated effects to decision-relevant populations for health technology assessment (HTA). We examine and formalize transportability in PAICs from an estimand-based perspective. We distinguish conditional and marginal treatment effect estimands and show how transportability depends on effect modification, collapsibility, and alignment between the scale of effect modification and the effect measure. Using illustrative examples, we demonstrate that even when effect modifiers are shared across treatments, marginal effects are generally population-dependent for commonly used non-collapsible measures, including hazard ratios and odds ratios. Conversely, collapsible and conditional effects defined on the linear predictor scale exhibit more favorable transportability properties. We further show that pairwise PAIC approaches typically identify effects defined in the comparator population and that applying these estimates to other populations entails an additional, often implicit, transport step requiring further assumptions. This has direct implications for HTA, where PAIC-derived effects are routinely applied within cost-effectiveness and decision models defined for different target populations. Our results clarify when applying PAIC-derived treatment effects to desired target populations is justified, when doing so requires additional assumptions, and when results should instead be interpreted as population-specific rather than decision-relevant, supporting more transparent and principled use of indirect evidence in HTA and related decision-making contexts.

Time-to-event endpoints are central to evaluating treatment efficacy across disease areas. In clinical trials with time-to-event endpoints, the information available for interim and final analyses is largely determined by the number of observed events rather than by the number of enrolled patients. Interim monitoring therefore requires assessing how many additional events are expected to accrue by scheduled future analysis dates. Quantifying uncertainty around these counts is essential for assessing whether planned information levels are likely to be reached, anticipating delays or event overrunning, and supporting operational decisions while the trial is ongoing. This is especially relevant in pediatric oncology trials, where event accrual is often uncertain. Although methods for predicting time to endpoint maturation are well established, interval prediction for event counts at fixed calendar times remains less developed. We propose a patient-level framework for constructing such intervals at interim analyses of time-to-event trials. Conditionally on the interim data, the future count follows a Poisson--binomial law with patient-specific event probabilities; we estimate this law using a conditional parametric bootstrap. Under standard regularity conditions, the bootstrap is consistent and yields asymptotically calibrated prediction intervals. The framework accommodates staggered entry, patient-level covariates, administrative censoring, random loss to follow-up, and possible dependence between entry dates and loss to follow-up before conditioning on the realised interim data. We study its operating characteristics in simulation studies and illustrate it using a real-world phase III trial in childhood acute lymphoblastic leukaemia.

Joint Species Distribution Models are essential for understanding how ecological covariates shape species communities. However, most existing approaches are limited by rigid parametric distributions for count data and the inability to model how interspecific associations change with those covariates. We introduce joint count transformation models, a novel framework designed to overcome these limitations. Our approach combines distribution-free marginal count transformation models for multiple species with a covariate-dependent latent Gaussian copula to model interspecific correlations, interpretable as Spearman's rank correlation on the observed count scale. All model parameters are estimated efficiently via joint maximum likelihood estimation, implemented in the R package tram. We apply this framework to model the joint abundance of three fish-eating bird species, using seasonality as the primary covariate. Our model successfully captured the complex, species-specific seasonal abundance patterns, including periods of high zero-counts and seasonal shifts in variance. Furthermore, the model revealed strong, seasonally-varying correlations between the species. These findings are consistent with an empirical approach and similar to those from the computationally expensive parametric Bayesian Hierarchical Modelling of Species Communities (HMSC) framework. Consistency, accuracy and feasibility of our approach are demonstrated in a simulation study for up to 10 species.

The Cox proportional hazards model is the most widely used regression model in univariate survival analysis, yet extensions to bivariate survival data remain scarce. We propose two novel extensions based on a Lehmann-type representation of the survival function. The first, the simple Lehmann model, is a direct extension that retains a straightforward structure. The second, the generalized Lehmann model, allows greater flexibility by incorporating three distinct regression parameters and includes the simple Lehmann model as a special case. The models admit a direct interpretation in terms of survival probabilities, providing a transparent, fully semiparametric framework for assessing covariate effects on both marginal survival probabilities and their dependence, without requiring specification of a copula or frailty distribution. To estimate the regression parameters, we build on a pseudo-observation-based approach for bivariate survival data and extend it to the generalized model via a two-step procedure. We establish consistency and asymptotic normality of the resulting estimators. The proposed approach is illustrated through simulation studies and an application to data from the Global Retinoblastoma Outcome Study.

Ordinal outcomes are common in clinical settings where they often represent increasing levels of disease progression or different levels of functional impairment. In this article, we focus on representing the average treatment effect for ordinal outcomes via intrinsic pairwise outcome comparisons captured through win estimands, such as the win ratio and win difference. Recognizing the value of baseline covariate adjustment toward enhanced precision, we first develop propensity score weighting estimators, including both inverse probability weighting (IPW) and overlap weighting (OW), tailored to estimating win estimands. Furthermore, we develop augmented weighting estimators that leverage an additional ordinal outcome regression to potentially improve efficiency over weighting alone. Leveraging the theory of U-statistics, we establish the asymptotic theory for all estimators, and derive closed-form variance estimators to support statistical inference. We also prove that all of the covariate-adjusted estimators do not compromise consistency for the target estimand even when the associated working models are incorrectly specified; hence these covariate-adjusted estimators are model-robust. Through simulations we demonstrate the enhanced efficiency of the weighted estimators over the unadjusted estimator, with the augmented weighting estimators showing a further improvement in efficiency except for extreme cases. Finally, we illustrate our proposed methods with the ORCHID trial, and implement our covariate adjustment methods in an R package winPSW.

Stochastic infectious disease models capture uncertainty in public health outcomes and have become increasingly popular in epidemiological practice. However, calibrating these models to observed data is challenging with existing methods for parameter estimation. Stochastic epidemic models are nonlinear dynamical systems with potentially large latent state spaces, resulting in computationally intractable likelihood densities. We develop an approach to calibrating complex epidemiological models to high-dimensional data using Neural Posterior Estimation, a novel technique for simulation-based inference. In NPE, a neural conditional density estimator trained on simulated data learns to "invert" a stochastic simulator, returning a parametric approximation to the posterior distribution. We introduce a stochastic, discrete-time Susceptible Infected (SI) model with heterogeneous transmission for healthcare-associated infections (HAIs). HAIs are a major burden on healthcare systems. They exhibit high rates of asymptotic carriage, making it difficult to estimate infection rates. Through extensive simulation experiments, we show that NPE produces accurate posterior estimates of infection rates with greater sample efficiency compared to Approximate Bayesian Computation (ABC). We then use NPE to fit our SI model to an outbreak of carbapenem-resistant Klebsiella pneumoniae in a long-term acute care facility, finding evidence of location-based heterogeneity in patient-to-patient transmission risk. We argue that our methodology can be fruitfully applied to a wide range of mechanistic transmission models and problems in the epidemiology of infectious disease.

Racial disproportionality in stop and search practices elicits substantial concerns about its societal and behavioral impacts. In London, Black individuals are about four times more likely to be stopped and searched than White individuals. Using data on stop and search events in London from January 2019 to December 2023, this paper aims to investigate disproportionality in the volume of stops for expressive crimes involving Black individuals compared to other ethnicities. We employ a semi-parametric partially linear structural regression method and introduce a Bayesian empirical likelihood procedure combined with double machine learning techniques to control for high-dimensional confounding and to accommodate strong prior assumptions. In addition, we show that the proposed procedure yields a valid posterior in terms of coverage. Applying this approach to the stop and search dataset, we find that racial disproportionality aimed at the Black community may be influenced by the borough racial composition when focusing on expressive crimes.

Distributionally robust optimisation (DRO) minimises the worst-case expected loss over an ambiguity set that can capture distributional shifts in out-of-sample environments. While Huber (linear-vacuous) contamination is a classical minimal-assumption model for an $\varepsilon$-fraction of arbitrary perturbations, including it in an ambiguity set can make the worst-case risk infinite and the DRO objective vacuous unless one imposes strong boundedness or support assumptions. We address these challenges by introducing bulk-calibrated credal ambiguity sets: we learn a high-mass bulk set from data while considering contamination inside the bulk and bounding the remaining tail contribution separately. This leads to a closed-form, finite $\mathrm{mean}+\sup$ robust objective and tractable linear or second-order cone programs for common losses and bulk geometries. Through this framework, we highlight and exploit the equivalence between the imprecise probability (IP) notion of upper expectation and the worst-case risk, demonstrating how IP credal sets translate into DRO objectives with interpretable tolerance levels. Experiments on heavy-tailed inventory control, geographically shifted house-price regression, and demographically shifted text classification show competitive robustness-accuracy trade-offs and efficient optimisation times, using Bayesian, frequentist, or empirical reference distributions.

Fairness audits are a key component of responsible machine-learning deployment. Yet, audit-recommendation reliability under incomplete protected-label access is still poorly understood. In this work, we focused on protected-label missingness in fairness mitigation audits. We introduced a seed-calibrated stress test to separate missingness effects from seed-to-seed movement already present under complete labels. Across ACS/Folktables tasks, missingness settings that retain some protected labels usually do not move selected mitigation methods beyond a complete-label seed-to-seed baseline. At $0%$ protected-label access, candidates collapse to an empirical-risk-minimization baseline and deterministic tie-breaking rather than revealing a broad missingness effect. We also found that threshold optimization can turn fairness gains on a single protected axis into intersectional harm above a seed baseline, and this threshold-optimizer finding persists under random-forest validation. Overall, our results highlight that protected-label missingness should be reported with seed-null calibration, candidate-set context, and intersectional consequences before it is treated as evidence of audit fragility.

Organizations increasingly deploy AI as a low-cost prediction layer in customer-facing decision processes, including demand sensing, service-quality monitoring, product testing, and market research, but AI-generated signals are unevenly reliable across tasks, products, and customer segments. Firms therefore still need scarce human validation (labels, audits, survey responses, or follow-up measurements) to anchor AI outputs to ground truth. Because human ground truth is itself noisy, varying across labelers and even across repeated judgments, the firm must collect and average several human labels per task, which makes human validation costly. We study how to allocate a limited human-validation budget across many AI-assisted tasks when reliability is heterogeneous and unknown before deployment. We cast this within tuned prediction-powered inference. Each human label both sharpens the AI-assisted estimate and reveals the task's rectification difficulty, the variance that remains after the AI prediction is optimally used as a control variate. If difficulties were known, the optimal allocation would follow a Neyman square-root rule; because they are unknown, we propose a policy based on upper confidence bounds that learns them online and steers validation toward tasks where AI is least reliable. We prove that the policy's terminal efficiency loss relative to the oracle allocation vanishes as the budget grows. In synthetic experiments and a real digital-twin survey with 68 tasks and over 2000 respondents, it closes most of the gap to the oracle when reliability is heterogeneous, outperforming uniform and epsilon-greedy allocation; on the survey data it also outperforms explore-then-commit pilot designs and cuts uniform's 10--12% gap to 2--6%. The value of AI depends not only on model accuracy but also on the operational policy that targets human oversight where AI errors matter most.

Structure-based and ligand-based computational drug design have traditionally relied on disjoint data sources and modeling assumptions, limiting their joint use at scale. In this work, we introduce Contrastive Geometric Learning for Unified Computational Drug Design (ConGLUDe), a single contrastive geometric model that unifies structure- and ligand-based training. ConGLUDe couples a geometric protein encoder that produces whole-protein representations and implicit embeddings of predicted binding sites with a fast ligand encoder, removing the need for predefined pockets. By aligning ligands with both global protein representations and multiple candidate binding sites through contrastive learning, ConGLUDe supports ligand-conditioned pocket prediction in addition to virtual screening and target fishing, while being trained jointly on protein-ligand complexes and large-scale bioactivity data. Across diverse benchmarks, ConGLUDe achieves competitive zero-shot virtual screening performance, substantially outperforms existing methods on a challenging target fishing task, and demonstrates state-of-the-art ligand-conditioned pocket selection. These results highlight the advantages of unified structure-ligand training and position ConGLUDe as a step toward general-purpose foundation models for drug discovery.

We present a suite of packages in R, Python, Julia, and C++ that efficiently solve the Sorted L-One Penalized Estimation (SLOPE) problem. The packages feature a highly efficient hybrid coordinate descent algorithm that fits generalized linear models (GLMs) and supports a variety of loss functions, including Gaussian, binomial, Poisson, and multinomial logistic regression. Our implementation is designed to be fast, memory-efficient, and flexible. The packages support a variety of data structures (dense, sparse, and out-of-memory matrices) and are designed to efficiently fit the full SLOPE path as well as handle cross-validation of SLOPE models, including the relaxed SLOPE. We present examples of how to use the packages and benchmarks that demonstrate the performance of the packages on both real and simulated data and show that our packages outperform existing implementations of SLOPE in terms of speed.

High-resolution estimates of population health indicators are critical for precision public health. We propose a method for high-resolution estimation that fuses distinct data sources: an unbiased, low-resolution data source (e.g. aggregated administrative data) and a potentially biased, high-resolution data source (e.g. individual-level online survey responses). We assume that the potentially biased, high-resolution data source is generated from the population under a model of sampling bias where observables can have arbitrary impact on the probability of response but the difference in the log probabilities of response between units with the same observables is linear in the difference between sufficient statistics of their observables and outcomes. Our data fusion method learns a distribution that is closest (in the sense of KL divergence) to the online survey distribution and consistent with the aggregated administrative data and our model of sampling bias. This approach significantly reduces bias in high-resolution estimates compared to baselines that rely on a single data source alone on a testbed that includes repeated measurements of three indicators measured by both the (online) Household Pulse Survey and ground-truth data sources at two geographic resolutions over the same time period.

A novel, stochastic approach to amputation, the process of introducing missing values to a complete dataset, is presented. It allows one to construct a wide variety of missingness patterns by only having to specify distributions of missingness indicators as opposed to specifying each missingness pattern manually. Missingness indicators are modeled in a principled way via copulas and Bernoulli margins, thus allowing one to incorporate dependence in missingness patterns. Besides more classical missingness mechanisms such as missing completely at random, missing at random, and missing not at random, the approach is able to model structured missingness such as block missingness and, via mixtures, monotone missingness, which are patterns of missing data frequently found in real-life datasets. Properties such as joint missingness probabilities or missingness correlation are derived mathematically. The flexibility of the approach in capturing different missingness patterns while only requiring to specify distributional assumptions on missingness indicators is demonstrated with mathematical examples and empirical illustrations in terms of a well-known example dataset of sufficiently small sample size that allows to identify each missing data point visually. Finally, an example application to multivariate financial time series is provided.

Ensuring robot safety can be challenging; user-defined constraints can miss edge cases, policies can become unsafe even when trained from safe data, and safety can be subjective. Thus, we learn about robot safety by showing policy trajectories to a human who flags unsafe behavior. From this binary feedback, we use the statistical method of conformal prediction to identify a region of states, potentially in learned latent space, guaranteed to contain a user-specified fraction of future policy errors. Our method is sample-efficient, as it builds on nearest neighbor classification and avoids withholding data as is common with conformal prediction. By alerting if the robot reaches the suspected unsafe region, we obtain a warning system that mimics the human's safety preferences with guaranteed miss rate. From video labeling, our system can detect when a quadcopter visuomotor policy will fail to steer through a designated gate. We present an approach for policy improvement by avoiding the suspected unsafe region. With it we improve a model predictive controller's safety, as shown in experimental testing with 30 quadcopter flights across 6 navigation tasks. Code and videos are provided.

Since the introduction of network psychometrics, several connections to statistical models in "classical" psychometrics (i.e., IRT, SEM, GLM) as well as to approaches from other research fields have been established. In this paper, these developments have been reviewed and synthesized and, based on an exploratory literature search, further advanced and presented in an accessible visual format. This perspective opens up promising opportunities to extend the psychometric-toolbox by incorporating and learning from statistical methodologies developed in other research domains, which often address similar or even identical problems. Highlighting these methodological commonalities may also foster collaboration across research fields that have traditionally remained largely independent. Moreover, awareness of these connections may render methodological development more systematic and goal-directed and may enable a meaningful division of labor, for example between the development of statistical methodology and its practical implementation for empirical research through software tools. Finally, these methodological advances provide new opportunities for empirical research and may contribute to a reconciliation with longstanding conceptual issues concerning psychometric constructs and, more broadly, psychological phenomena.