Papers

Existing methods for identifying rapid-responding behaviour in large-scale assessments require parametric assumptions about the population. In this study, we propose a novel, non-parametric, mutual information-based framework of methods as an alternative. The methods within this framework compute the mutual information of the observed responses and discretised response times and maximise the information gain to determine a threshold that differentiates rapid responses from engaged responses. The only difference between the methods is the number of categories in the relevant variables. We present three methods explicitly. The first method uses response correctness and binarised response times. The second method uses correctness and categorises time into three groups. The third method uses raw responses and binarised times. We applied these methods to mathematics achievement data collected through the Programme for International Student Assessment in 2022. Furthermore, we examined the behaviour and usability of the first method in certain realistic conditions at the population and realised levels. The results indicated that the proposed framework is a viable alternative for identifying rapid responses. The framework's novelty lies in its non-parametric nature and its ability to utilise raw responses instead of correctness. Finally, we discuss some possible future research directions on this topic.

Matrix graphical models aim to characterize conditional dependence structures in matrix-variate data under a separable covariance assumption. In this framework, the precision matrix is decomposed as a Kronecker product, enabling separate modeling of undirected graphs across row and column domains. Existing methods have been developed for this problem, including likelihood-based approaches and regression-based procedures for graph estimation. Likelihood-based methods estimate precision matrices directly and recover graph structures indirectly, whereas regression-based approaches directly target estimating edges among variables, thus outperforming the former. However, existing regression-based methods are based on multiple penalized regression problems, which naturally yields asymmetry in estimated graphs and computational difficulty in selecting tuning parameters. To address the limitations, we propose a joint estimation of partial correlations in matrix graphical models. The proposed method estimates all partial correlations simultaneously within a unified optimization framework, thereby preserving symmetry and easing the pain of selecting the best models. Numerical studies demonstrate that the proposed method improves graph recovery performance compared to existing approaches. We also analyze protein expression data collected from patients with pulmonary tuberculosis, measured repeatedly at multiple time points, where the proposed method compares protein networks between two groups of patients and recovers the temporal dependence structure.

Large Language Models (LLMs) have achieved remarkable reasoning capabilities by utilizing chain-of-thought (CoT) as a scratchpad for intermediate stages of thinking. However, CoT techniques require explicit supervision on thinking tokens, which requires rich, task-specific data. In this work, we propose Abstract Token Curriculum (ATC), a novel curriculum learning framework that elicits effective continuous intermediate representations without direct supervision or manual scratchpad design. ATC gradually increases problem complexity through a sequence of distributions, training the model to develop internal abstract “thoughts” in the continuous representation space. This paper provides both theoretical and experimental evidence for the benefits of ATC and its advantages over previous methods for training continuous thoughts. Theoretically, we show that for learning parity functions with single-layer softmax attention using ATC, attention naturally focuses on the CoT tokens in the context that provide the “easiest path” to predicting the next token. Experimentally, we show ATC's effectiveness on graph reachability and arithmetic learning tasks.

Several institutions test one family of hypotheses under family-wise error control but cannot pool their data, so each site releases only a quantized report of its p-value for each hypothesis, such as whether the p-value is significant. The report's resolution is the number of values it can take. We quantify the power lost to such reports relative to the most powerful centralized procedure given all p-values (the oracle), with reports and combining rule chosen optimally. A report from which the p-value can be recovered loses no power, so the loss is due to compression, not decentralization. Under the stated regularity and power-objective conditions, every finite resolution loses power, and the optimal loss decays as the inverse square of the resolution: equal-width intervals attain this order, and no partition into as many cells improves it. In two-site Beta examples, optimized one-bit reports are nearly lossless for a single hypothesis but lose several times as much power for two. At level 0.05 with two Beta(1,2) hypotheses, optimally combined three-bit equal-width reports retain 99.2% of oracle power. We give the power-optimal combining rule for given reports, and a rule using only their exactly known null distribution, valid under arbitrary dependence across hypotheses.

We study sparse composite quantile regression (CQR) for distributed data with heterogeneous honest sites and Byzantine workers. Honest sites share a common slope but may differ in their covariate distributions, error laws, and quantile intercepts. The proposed heterogeneity-calibrated robust CQR (HC-RCQR) profiles local intercepts and calibrates scores using an approximate inverse profile Hessian. Honest workers transmit the resulting vectors, whereas Byzantine workers may send arbitrary vectors. The server updates the estimate by coordinatewise trimming and soft thresholding. A scalar example shows how unequal honest-site curvatures allow intermediate Byzantine reports to survive trimming and how ideal calibration reduces their possible effect. We also establish nonidentification of the mean of unrestricted honest-site slopes when fault identities are unknown. Under suitable conditions, we establish conditional contraction and support-recovery guarantees. The bound separates score offset, sampling fluctuation, contamination, calibration error, and the Newton remainder. Simulations and a bike-demand study examine performance under heterogeneous data and adversarial messages.

Peptide-HLA binding prediction is a critical step in neoantigen identification for personalized cancer immunotherapy and holds significant clinical value. However, the training data available for many HLA alleles are extremely limited, which severely constrains the performance of conventional methods on this task. Parameterized quantum circuits are hypothesized to induce inductive biases beneficial for learning from small datasets, yet their application to biological sequence prediction remains underexplored. To address this, we propose a hybrid quantum-classical neural network (HQNN) specifically designed for peptide-HLA binding prediction. HQNN integrates multi-source biological feature encoding with parallel quantum feature extractors and a quantum-enhanced classifier. On two HLA alleles (A*02:01 and B*07:02), HQNN outperforms a parameter-matched classical CNN baseline across all training sizes, with the performance gap widening as training data decreases. Ablation studies confirm the respective contributions of the quantum feature extraction module and the quantum classifier. In noise-aware simulations, performance degrades only mildly, and such degradation is reasonable and acceptable under realistic quantum hardware noise levels. These results suggest that hybrid quantum-classical architectures can provide practical sample-efficiency gains for immunoinformatics tasks in low-data regimes.

Human pairwise comparisons provide a reference for evaluating large language models (LLMs), but collecting sufficient judgments for each new release is costly and time-consuming. LLM judges offer a scalable alternative, although their comparisons may differ systematically from human preferences and across judges. We study the ranking of a new model that has received LLM-judge comparisons but no human comparisons. We propose ANCHOR (ANchored Comparisons for Human-reference inference with Orthogonal Riesz correction), which uses historical human and LLM comparisons to learn judge-specific sensitivities to human score differences and feature-dependent judge biases. These estimates are then used to infer the new model's human-reference score from its judge comparisons. The framework allows the feature distribution to change between historical and new-model comparisons. For inference, we construct a Neyman-orthogonal estimator through a joint Riesz correction that removes the first-order effects of estimating the historical human scores, judge sensitivities, and bias functions. We establish identification, convergence rates, and asymptotic normality with consistently estimable variance, and show that ANCHOR attains the semiparametric efficiency bound. Simulations demonstrate gains in score estimation and ranking accuracy. On Chatbot Arena, ANCHOR achieves the lowest score RMSE and insertion MAE among competing methods, with narrower score intervals on average.

Growth curve models are widely used in psychological research, and variable selection can help identify baseline characteristics associated with longitudinal heterogeneity. However, conventional inference after data-driven variable selection can be invalid because the same outcome data are used for both selection and inference. We develop a data-fission framework for post-selection inference in growth-curve models that separates the information used for selection and inference while retaining all participants in both stages through the addition and subtraction of Gaussian noise. The framework accommodates flexible variable-selection procedures and targets covariance-weighted linear projection parameters in the selected working model. It provides exact inference when the covariance structure is known, and we establish asymptotic validity under suitable regularity conditions when the covariance structure is estimated. Simulations show that the proposed method provides valid inference and improves efficiency over subject-level data splitting, whereas naïve post-selection inference can be biased. An application to the Longitudinal Study of American Youth illustrates the proposed framework.

This paper establishes a unified non-linear regularization framework for high-dimensional $M$-estimation, encompassing both linear models and Cox's proportional hazards models. Rather than relying on traditional additive non-convex penalties, the proposed paradigm embeds sparsity directly into the transformation for the physical parameter $\theta=\phi(\beta)$ using a smooth ($C^2$) component-wise "ReParametrization map for Sparsity (RePS)" $\phi$, and the penalty term is $\lambda \Vert \beta \Vert_1$ rather than $\lambda \Vert \theta \Vert_1$. This structural formulation dynamically adapts to local parameter scales, suppressing high-dimensional noise while simultaneously recovering unbiased oracle asymptotic normality under the large-sample limit. Through the Primal-Dual Witness method, we establish a unified oracle equivalence for both model classes under a liberated micro-penalty scaling regime $\lambda \ll n^{-1/2}$, accommodating model-specific structures such as the shift-invariance in survival analysis. Extensive Monte Carlo simulations demonstrate that the proposed framework consistently achieves superior false-positive control and high 95% confidence interval coverage in high-dimensional linear models, while also delivering performance comparable or superior in all aspects to state-of-the-art methods like SCAD and MCP in proportional hazards models.

Materials discovery and design campaigns can be formulated as constrained multi-objective Bayesian optimization (CMOBO) problems, within which each experimental decision negotiates between two coupled but competing goals: discovering feasible candidates and refining the underlying Pareto front. Here we recast acquisition-function choice as an adaptive policy-selection problem over a portfolio of conventional and feasibility-focused acquisition functions. This was done using two controllers: UCB-Bandit, a modified UCB multi-armed bandit, and Agentic-Switch, a multi-agent decision system driven by a large language model (LLM). Both were evaluated against fixed-policy baselines in silico across five synthetic benchmark functions and two materials design case studies. The adaptive policies performed competitively in terms of both cumulative feasibility count and feasible hypervolume improvement, while each individual acquisition function performed well for only one metric, suggesting that adaptive policies are better suited for constrained materials science problems.

Multivariate longitudinal data may exhibit non-Gaussian margins, nonlinear dynamics, and response vectors with composition that varies across waves. To account for these features, we introduce a vector drawable vine (VD-vine) copula that extends conventional drawable vine copulas from scalar to vector-valued nodes. Here, the response vector at each wave forms a multivariate marginal, and serial dependence is captured through a sequence of linking vector copulas. We establish that the VD-vine is itself a vector copula and reduces to a conventional drawable vine for scalar nodes. Recursive forward and backward conditional transports are derived that enable efficient likelihood evaluation and predictive simulation, with parsimonious reductions under finite-order Markov and stationary restrictions. Unconstrained parameterizations for Gaussian and FGM linking vector copulas, flexible multivariate marginals, and Bayesian variational inference provide a practical implementation. Simulations show improved predictive accuracy when the marginals are asymmetric and serial dependence is multivariate, with little loss under a correctly specified Gaussian panel vector autoregression. In an eight-wave Australian panel of 1,093 individuals with varying response vectors, the full VD-vine delivers the best cross-validated distributional forecasts among the models considered, establishing the benefit of capturing asymmetry and nonlinear dependence.

Active subspace methods provide a framework for quantifying predictive uncertainty in high-dimensional models by identifying and performing inference along parameter directions that have the greatest influence on the model output. However, the construction of active subspaces requires storing many full-dimensional model gradients, which becomes prohibitive as model size increases. We address this limitation by proposing Compressed Active Subspaces (CAS), a scalable approach that first maps the model parameters to a compressed space using a structured isometric embedding and then constructs the active subspace within this reduced parameterization. Our approach substantially reduces the memory required for active subspace construction and enables Bayesian inference for large models where standard active subspace methods become impractical. We demonstrate the scalability of CAS on neural networks of increasing size while maintaining predictive performance and robust uncertainty estimates.

Suppose we observe the first $n$ points of a sequence of random variables having length $n+1$, and wish to estimate a functional of the unobserved final point and the empirical measure of the $n$ observed training points. Such next-token functionals include the probability that the next token is novel (also known as the surprise probability), the tail probability of the minimum distance between the next token and training points, and the test error of a classifier trained on the observed points. All of these quantities are classically estimated by the leave-one-out method, which is inconsistent under temporal dependence. We propose a leave-a-window-out estimator, which deletes a window of length $\tau$ after each index before forming the empirical measure and reduces to leave-one-out at $\tau = 1$. Under natural assumptions, we show that the error of our estimator decays at a parametric rate for any stationary $\beta$-mixing process that also admits a Marton coupling. Our results thus cover several natural functionals on a large class of stochastic processes. We complement these upper bounds with a sharp minimax lower bound for estimating the surprise probability on mixing Markov chains. Simulations on Markov chains, moving-average processes, and autoregressive processes show that our estimator succeeds in many scenarios where leave-one-out and add-constant baselines fail.

Clinical B-mode images are widely available as potential data sources for quantitative ultrasound (QUS) analysis for tissue characterization. However, standard clinical ultrasound devices apply unknown log-compression to RF envelope data before display and storage. Previous work has demonstrated estimation of the underlying RF envelope statistics in the presence of an unknown compression law. Using Fisher information analysis, we show that finite-offset log compression causes severe information loss when estimating the Rayleigh scale $\sigma$, which controls diffuse speckle. For a single image window, unknown compression raises the minimum achievable variance for unbiased estimation of $\sigma$ by a compression-independent factor of approximately 34.98. When $M$ equal-sized windows share the same unknown compression settings, the excess variance decays as $1/M$; even in the most favorable regime, reducing the variance inflation factor below 1.1 requires 340 windows. Our analysis treats the contrast parameter $a$ as unknown and the boundary offset $b$ as known; estimating $b$ experimentally shows even larger variance. We validate this theory using synthetic estimation experiments and demonstrate RF-scale recovery on real RF-envelope windows from the OASBUD dataset. Together, these results clarify the limitations of using routine B-mode images for QUS.

Receiver Operating Characteristic (ROC) and Precision-Recall (PR) curves are widely used to assess the discrimination ability of forecasts for binary events, such as threshold exceedances or warnings of extreme events. In weather forecasting, forecasts are provided as spatial fields, yielding location-wise ROC and PR curves that are often aggregated to facilitate comparison. However, the effect of the aggregation strategy on performance assessment remains poorly understood. We investigate how different aggregation strategies for ROC and PR curves affect the assessment of discrimination ability. In particular, we identify conditions under which aggregation strategies satisfy two desirable properties for fair comparison: preservation of dominance between forecasts and preservation of concavity or achievability of the curves. We obtain sufficient conditions and propose two strategies satisfying them. They are compared with existing strategies from the literature, and we analyze their properties and highlight potential pitfalls that may lead to misleading interpretations. Based on these findings, we provide practical guidelines for the interpretation of aggregated ROC and PR curves. The proposed framework is illustrated with AI-based global weather forecasts, showing how different aggregation strategies can yield different rankings of competing forecasts.

Actor-partner interdependence models (APIMs) describe how one member's exposure is associated with that member's and the partner's outcomes, but actor and partner coefficients are causal only under additional design and identification assumptions. We formulate APIM-style questions as dyadic target trials under within-dyad interference and make three links between the associational and causal formulations explicit. First, under joint exchangeability, positivity, and a homogeneous correctly specified treatment-related conditional mean, APIM actor, partner, interaction, and joint coefficients equal the corresponding causal contrasts. Second, for binary dyadic treatment, multiplying the true member-specific propensities induces an exact one-parameter discrepancy delta(H) = Cov(A_1, A_2 | H): the four joint-cell errors are (+delta, -delta, -delta, +delta), total variation is 2|delta|, and IPW contrast bias has a closed-form representation. Third, the same discrepancy affects AIPW through its product with outcome-regression error. A 36-scenario simulation and a true-marginal factorization diagnostic show that large separate-propensity bias under residual dependence is primarily factorization bias, while flexible AIPW reduces nonlinear-model bias but not weak-support instability. In the strongest dependence condition, factorizing even the true marginals reversed the spillover contrast and inflated the interaction to nearly five times its true value. We apply the framework to 2,400 Health and Retirement Study couples studying retirement transitions and later depressive symptoms. Joint retirement was rare, residual treatment dependence was substantial, and estimates were sensitive to support and propensity flooring. The contribution is a design-and-diagnostic bridge from APIM questions to causal dyadic analyses, not a new general interference or AIPW theory.

In infectious disease transmission, regression and variable selection for time-to-event outcomes are complicated by transmission-induced dependence that violates standard independence assumptions. Pairwise survival analysis addresses this dependence by modeling contact intervals in ordered pairs, defined as the time from the onset of infectiousness in the pair's first individual to their first infectious contact with the second individual. We propose a penalized pairwise accelerated failure time model for contact intervals that accommodates simultaneous covariate effects on infectiousness and susceptibility while jointly representing within-population transmission and infection from external sources. For variable selection with many candidate covariates, we develop adaptive LASSO penalized maximum likelihood estimation with a coordinate descent algorithm, treating distributional baseline parameters as unpenalized nuisance parameters. We establish selection consistency and asymptotic normality of the adaptive LASSO estimator under profile-likelihood regularity conditions in a local asymptotic regime in which information accumulates over all ordered pairs that are ever at risk of transmission during follow-up. Simulation studies show accurate variable selection and improved estimation accuracy compared with the unpenalized estimator and the LASSO. We illustrate the method using 2009 household influenza A (H1N1) surveillance data from Los Angeles County, where the adaptive LASSO produces a sparse, interpretable model and improves the precision of selected covariate effect estimates, including the effect of antiviral prophylaxis on susceptibility.

Joint-embedding predictive architectures (JEPA) have emerged as a promising self-supervised pretraining paradigm for time series, learning representations by predicting target embeddings in latent space rather than reconstructing raw signals. Yet evidence on their benefits remains mixed, and most studies test only a single backbone or a narrow set of architectures, leaving unclear whether JEPA pretraining is a reliable improvement or one that depends heavily on the downstream model. We address this gap through a large scale evaluation of one JEPA instantiation across nine backbones and eleven benchmarks spanning temporal and spatio-temporal forecasting, the most extensive cross architecture assessment of JEPA for time series to date. We find that the benefit of this instantiation varies sharply across backbones, producing consistent gains for some architectures and consistent degradation for others, even on the same dataset. This pattern holds across both task families, indicating the variability is a general property of this instantiation rather than a dataset specific artifact worth accounting for when choosing a backbone in practice.

This tutorial provides a unified view of the emerging area of bilinear learning and control. Using linear systems as a benchmark, it explains what fundamentally changes in the bilinear settings, how recent theory addresses finite-sample learning and control, and how these ideas connect to broader themes in nonlinear control, representation learning, and data-driven decision making. For learning, we emphasize tools that are particularly useful in the bilinear settings, such as one-sided Bernstein's inequality for dependent and heavy-tailed covariates, blocking arguments, and martingale concentration for input-dependent noise. We then apply these tools to obtain finite-sample learning guarantees for fully observed bilinear systems, partially observed bilinear systems, and linear systems with bilinear observations. For control, we discuss quadratic control from bilinear observations, where the classical separation principle fails, and review tractable approaches based on belief-space receding horizon control. We also cover stabilization of bilinear dynamics under state feedback using semi-definite programming, LMI relaxations, sum-of-squares methods, and Koopman-based lifting. We conclude by discussing connections to reinforcement learning and machine learning, and some open problems in combined learning and control of bilinear systems.

Despite the widespread use and success of generative AI techniques today, theoretical guarantees on learning a distribution supported in $d$ dimensions from $n$ samples degrade as $O(n^{-1/\Theta(d)})$, though shown to be minimax optimal. We hypothesize that present bounds are too pessimistic because smoothness assumptions are not enough to capture the structure of distributions that often appear in real applications. Consequently, we introduce the class of sparse priors and define the "Sparse Dimension" as a measure of sparsity of a prior over the space of all distributions. We show that distribution learning under a $k$-sparse prior achieves a Bayesian risk lower bound of $\Omega(\sqrt{k/n})$ under common distance metrics, and show a matching (up to logarithmic terms asymptotically in $n,k$) upper bound for the TV distance under mild additional assumptions. We show the statistical equivalence of distribution learning and learning to sample in the Bayesian setting so that our results apply to learning to sample as well. While $k$ can still depend on the dimension $d$, or a notion of intrinsic dimension, our results show that learning under an appropriate prior overcomes the curse of dimensionality with respect to the dependence on $n$.

Feature importance is central to interpretable machine learning, but the term "importance" encompasses several fundamentally different notions of relevance. We develop a unified perspective based on null importance: a population-level characterization of when a feature is irrelevant under a specified notion of relevance. We consider standard notions of null importance arising from marginal and conditional statistical relevance, predictive risk, functional invariance, and causal effects, and show how these notions answer different scientific questions. We illustrate the framework in two applications in which the distinction is particularly consequential: algorithmic fairness, where common fairness criteria correspond to different notions of null importance, and genomic perturbation modeling, where different notions of relevance lead to different conclusions about what a prediction model has learned. The framework connects three aspects of feature analysis: the scientific question defining relevance, the data and model assumptions that shape how different null notions relate, and the methods used to assess importance. We establish sufficient conditions under which null notions coincide and give counterexamples showing how they diverge when those conditions fail. We then characterize which nulls different method families target and when their zero-importance statistics identify those targets. Finally, simulations spanning feature dependence, redundancy, nonlinearity, hidden features and other standard phenomena, along with case studies on image and multiomics data, provide empirical evidence for these theoretical distinctions and their practical consequences. Taken together, these results provide a common statistical language for relating scientific questions, data-generating assumptions, and algorithms, and clarify the conclusions that feature-importance analyses can support.

To design many Bayesian studies, the sample size is chosen to attain sufficient power to reject a null hypothesis or a desired probability that an interval estimate is sufficiently narrow. Determining the minimum suitable sample size for studies with complex models typically requires intensive Monte Carlo simulation. One simulation-based strategy to consider the sample-size space involves estimating sampling distributions of posterior summaries, such as posterior probabilities or interval estimate lengths, at various sample sizes. Another strategy instead explores these sampling distributions via stochastic root finding over the sample-size space, where the entire sampling distribution of posterior summaries is not estimated at any given sample size. The performance of these competing approaches can be assessed using their Monte Carlo sampling distributions of sample size recommendations across repeated implementations. In this paper, we theoretically derive these Monte Carlo sampling distributions for various Bayesian sample size determination approaches. To promote best practices for Bayesian sample size determination, we conduct extensive numerical studies to estimate these Monte Carlo sampling distributions for the sampling distribution estimation and exploration approaches to assess their bias, precision, and sensitivity to tuning parameters.

Antimicrobial resistance is expected to claim 10 million lives per year by 2050, and resource-limited regions are most affected. Raman spectroscopy is a novel pathogen diagnostic approach promising rapid and portable antibiotic resistance testing within a few hours, compared to days when using gold standard methods. However, current algorithms for Raman spectra analysis 1) are unable to generalize well on limited datasets across diverse patient populations and 2) require increased complexity due to the necessity of non-trivial pre-processing steps, such as feature extraction, which are essential to mitigate the low-quality nature of Raman spectral data. In this work, we address these limitations using Sharpness-Aware Minimization (SAM) to enhance model generalization across a diverse array of hyperparameters in clinical bacterial isolate classification tasks. We demonstrate that SAM achieves accuracy improvements of up to 10.5% on a single split, and an increase in average accuracy of 2.7% across all splits in spectral classification tasks over the traditional optimizer, Adam. These results display the capability of SAM to advance the clinical application of AI-powered Raman spectroscopy tools.

While kernel-based learning methods for scalar-valued functions have been extensively studied, comparatively little work considers nonlinear vector-valued operators. Such operators with vector-valued outputs, e.g., state vectors and trajectories, commonly arise in system identification problems. In this paper, we study the problem of learning a vector-valued operator under the framework of vector-valued kernel ridge regression. We assume that the outputs take values in an arbitrary Hilbert space and are corrupted by additive noise with bounded norm. Under the assumption that the unknown operator belongs to a vector-valued reproducing kernel Hilbert space, we derive a pointwise deterministic bound on the discrepancy between the kernel ridge regression estimator and the ground truth in the corresponding Hilbert space norm. Numerical simulations under two different scenarios are presented to validate the theoretical results.

In evidence-based policymaking, typically one experimental sample is observed, then a learned policy recommendation is implemented at scale. Policies learned from the experimental data can perform well in expected welfare, yet random sampling in the experiment can produce recommendations with poor welfare outcomes. In this paper, we ask: how should policy learning algorithms balance expected welfare against sampling risk? Our main contribution is to show that algorithmic stability plays a central role in characterizing and navigating the tradeoff. Intuitively, if a policy learning algorithm's recommendation remains stable when one experimental unit is replaced, then that algorithm has limited sampling risk. We propose a method for policy learning called policy-vote bagging, which learns treatment decisions on many subsamples then averages their votes into treatment probabilities. Relative to using one subsample, averaging across subsamples preserves expected welfare and improves expected utility for a risk-averse researcher. We derive sharp bounds linking estimation accuracy, subsample size, and welfare variation, including an exact guarantee under CARA utility.

We study the omitted variable bias (OVB) problem in canonical difference-in-differences (DiD) designs when unobserved confounding induces departures from the parallel trends assumption. Our results provide a novel characterization of the OVB formula for the average treatment effect on the treated (ATT), which is of independent interest. We show how the ATT bias is mainly governed by the strength of confounding in the treatment assignment mechanism and provide alternative ways of quantifying this strength, such as (i) changes in the average odds of treatment among the treated, (ii) confounding imbalance between treated and control units, or (iii) variation explained in treatment odds among the untreated. Building on these results, we offer sensitivity statistics for routine reporting, describing the minimum strength of confounding required to overturn the conclusions of a DiD study, as well as formal bounds on the strength of confounders based on comparisons to observed covariates or pre-trends. Finally, we provide flexible and efficient statistical inference methods for the bounds on ATT, which can leverage modern machine learning algorithms for estimation. We demonstrate the utility of our approach in an empirical example that estimates the effects of minimum wage on teen employment.

PURPOSE: Dynamic Bayesian borrowing methods adapt the contribution of external information according to agreement or disagreement with current data. Qualify-Then-Borrow (QTB) makes explicit how scientific source assessment determines what external information reaches the borrowing model and when observed outcome agreement should determine borrowing strength. METHODS: QTB uses five steps: define the target; assess the external data; classify them as qualified, repairable, or not qualified; use them directly, repair them before use, or exclude them; and then apply the selected Bayesian borrowing method. Dynamic borrowing is the final step and operates on residual empirical compatibility after identified material differences have been addressed. We evaluated 82 prespecified scenarios with 10,000 replications each. RESULTS: Outcome data alone could not determine the appropriate borrowing decision. Two settings with the same external outcome distribution produced similar robust-mixture behavior but different QTB decisions because one had a known endpoint problem. When misclassified data were borrowed, full pooling reduced 95% coverage to 60.8%; robust-mixture downweighting improved coverage to 91.8% but left a -4.8 percentage-point bias. Repair of a measured population shift reduced absolute bias from 1.94 to 0.80 percentage points. Under strong negative residual drift, Type I error reached 10.0% despite downweighting. With fully compatible data, QTB reduced to the usual robust-mixture analysis. CONCLUSION: QTB is a five-step decision architecture, not a new prior or compatibility statistic. It determines whether external information proceeds unchanged, requires repair, or is excluded before dynamic borrowing operates on residual empirical compatibility. Qualification does not certify exchangeability, and downweighting is a safeguard rather than a guarantee against residual bias.

In clinical settings, predicting binary outcomes is often complicated by right-censoring, which prevents us from distinguishing individuals in whom the event never occurs (i.e., negative, or non-susceptible) from those where it occurs after the censoring time (i.e., positive, or susceptible). To discriminate between these subpopulations, we can use mixture cure models (MCMs), which model both the outcome probability over an infinite horizon and its time-to-event (TTE) distribution when observed. However, the TTE approach requires event timestamps, which can be unreliable, costly to obtain, or biased due to disparities in clinical care. Alternatively, the task can be formulated as positive-unlabeled (PU) learning, in which individuals without an observed event are grouped as "unlabeled" rather than negative, recognizing that this group includes positives whose event occurred after censoring. Here we analyze the relationship between these frameworks and provide evidence-based recommendations for choosing between them. We begin by showing that both families optimize a shared likelihood but differ in how they constrain the probability of event occurrence prior to censoring among susceptible patients, which in PU learning is known as the labeling propensity; and that an MCM may be viewed as a PU model whose labeling propensity is parameterized as a TTE distribution, which makes the model identifiable. In systematic simulations and two real-world clinical cohorts, we examine whether and when including the TTE component benefits binary classification. We show that MCMs are preferred in only one specific regime: under selective censoring where the exact features governing censoring cannot be identified. Otherwise, PU learning achieves equivalent performance while avoiding challenges and potential biases associated with the TTE approach.

We propose a framework for restricted inference on random dot product graphs whose latent positions lie on an unknown low-dimensional support manifold. For general decision problems, we propose semisupervised decision rules that use auxiliary data to learn the support manifold. Specifically, our rules use the Isomap manifold learning procedure to construct a low-dimensional Euclidean representation of the observed graph, in which space an isometrically invariant function maps configurations of points to actions. We study the behavior of the proposed rules as the quantity of auxiliary data sampled from the unknown support manifold increases. We show that, as the auxiliary sample size increases, the risk of the semisupervised rule converges to the risk of an oracle rule that relies on the maximal amount of low-dimensional Euclidean structure that can be extracted from the support manifold. Examples, applications, and simulation studies are deferred to a sequel.

We propose a two-view weighted-concordance operator (WCO) framework for offline change-point detection in weakly dependent text networks. Weighted word-co-occurrence graphs yield first-order direct-co-occurrence and second-order shared-context views, projected into paired Euclidean observations using maps learned from an independent pilot corpus. A WCO scan detects changes in cross-view dependence. A dependent multiplier bootstrap tests for a change, while a segment-wise block bootstrap gives a descriptive stability interval for its location. We establish null weak convergence, bootstrap validity, asymptotic size control, consistent localization under a single identifiable change, and perturbation bounds for token-level errors. CUSUM and Gaussian-kernel MMD scans provide complementary benchmarks. In the 1915-1921 New Youth corpus, the method detects a change in November 1919, with a descriptive 95 percent stability interval from April 1919 to June 1920, consistent with the linguistic transition surrounding China's May Fourth and New Culture movements.