multivariate dependence modeling

This study investigates the predictability of dependence structures in conditional copulas while accounting for structural breaks in time series. To this end, the authors propose a robust score-type test that does not require prespecifying a parametric copula family. By integrating distributional regression with a local Gaussian copula approximation, the method flexibly captures dynamic dependence patterns and accommodates complex marginal dynamics. The approach is formulated within a semiparametric framework and employs a multi-stage estimator, with inference conducted via the moving block bootstrap. The asymptotic distribution of the test statistic is rigorously derived, and both Monte Carlo simulations and empirical analyses demonstrate its strong finite-sample performance.

This study addresses the limitations of conventional econometric inference, which often relies on pre-specified clustering, factor, or sparsity assumptions about dependence structures without data-driven validation. The authors model diverse dependence structures as covariance geometries in a Hilbert space and construct low-dimensional “dependence profiles” via projection-based similarity scores, enabling data-adaptive learning of dependence. They establish identifiability conditions for these profiles and derive finite-sample classification error bounds, revealing that indistinguishability arises from overlapping tangent spaces of the covariance geometries. Building on this insight, they propose an oracle-adaptive inference framework guided by dependence profiles, which yields consistent and asymptotically normal estimators of the underlying dependence structure and achieves inference performance equivalent to that of an oracle with prior knowledge of the optimal structure.

Traditional Spearman’s ρ fails to capture non-monotonic dependence structures and is sensitive to marginal distributions. Method: We propose a margin-free, unit-square-based generalized Spearman correlation coefficient, constructed within the Hilbert space of square-integrable functions. Marginal invariance is achieved via uniformity-preserving transformations and copula modeling; a novel randomized inverse transformation generates extremal singular copulas, enabling a parametric copula family that continuously interpolates non-monotonic dependence strength and supports symmetry detection. Contributions/Results: Leveraging orthogonal expansions in Legendre polynomials and cosine bases, we derive tight analytical bounds. The sample estimator is shown to be uniformly consistent and asymptotically normal. Empirical evaluations demonstrate superior performance in exploratory dependence analysis, symmetry identification, and non-monotonic density modeling compared to existing measures.

This work addresses the longstanding limitation in conditional density estimation—namely, the absence of closed-form solutions for multivariate conditional densities under non-Gaussian assumptions. We propose a generative conditional density estimation framework grounded in copula modeling and analytic conditionalization in latent space. Methodologically, we first establish the inheritability of “conditional stability” under mixture and transformation operations, thereby extending analytically tractable conditional families to non-Gaussian, nonlinear, and cross-dimensional settings. The core components include a Gaussian Mixture Copula Model (GMCM), an explicit latent-space conditionalization mechanism, and joint copula modeling. Experiments on synthetic and real-world datasets demonstrate substantial improvements in conditional density estimation accuracy and robustness to missing data imputation. Crucially, our approach enables efficient, differentiable, and sampling-free deterministic conditional inference.

This paper addresses the challenge of modeling nonlinear and asymmetric dynamic relationships among macroeconomic and financial variables. We propose the first scenario-analysis-oriented, dynamic nonparametric multivariate Bayesian machine learning framework. Methodologically, we adapt classical econometric tools—including conditional forecasting and generalized impulse response analysis—to high-dimensional Bayesian nonparametric models, integrating dynamic factor extensions and Monte Carlo simulation to enable asymmetric shock response estimation and conditional scenario inference. Our key contribution is the first systematic integration of traditional scenario-analysis tools with nonlinear Bayesian machine learning, explicitly capturing structural asymmetry. The framework is validated across three empirical domains: financial stress testing, macroeconomic risk assessment, and cross-border spillover analysis. Results demonstrate substantial improvements in risk measurement accuracy and cross-jurisdictional early-warning capability, offering a novel paradigm for prudential regulation and policy evaluation.

This study addresses the challenge of effectively characterizing nonlinear statistical dependence between two random variables while controlling for the influence of covariates. To this end, it proposes partial copula as a theoretical framework for nonlinear partial correlation, extending the classical notion of linear partial correlation to more general nonlinear settings. The work establishes a formal connection between partial copulas and conditional copula dependence structures. Through rigorous theoretical analysis and simulation experiments, the study demonstrates that partial copulas accurately capture dependence relationships after adjusting for covariates. Moreover, it reveals their potential in causal inference for identifying the sign of causal effects, thereby offering a novel tool for nonlinear causal discovery.

This study addresses the challenge of effectively disentangling and quantifying the individual contributions of marginal effects and dependence structure in multivariate statistical functionals under arbitrary marginal distributions, while establishing sharp extremal bounds when the dependence structure is uncertain. To this end, the authors propose a conditional copula–based decomposition that unifies the representation of multivariate functional expectations as combinations of marginal distributions and conditional copulas, thereby constructing a quantile–copula analytical framework. By introducing Δ-antitonicity to characterize the class of functionals amenable to sharp bounds, and integrating concordance order with functional integration techniques, the work extends the applicability of copula theory in uncertainty modeling. The framework’s generality and the tightness of its extremal bounds are demonstrated through applications in risk measures, stochastic dominance probabilities, information entropy, and option pricing.

This study addresses the computational challenges and intractable model space inherent in high-dimensional Bayesian vine copula model selection. The authors propose a novel framework that integrates loss-driven priors with shotgun stochastic search to simultaneously identify vine structures, select copula families, and estimate parameters. By employing a loss-guided sparse prior to shrink the model space and leveraging an efficient random search strategy, the method substantially enhances computational feasibility and efficiency in high-dimensional settings. Empirical evaluations on both simulated data and real-world ETF asset returns demonstrate that the proposed approach achieves estimation accuracy comparable to existing methods while offering markedly superior computational performance.

This work addresses the challenges of joint distribution modeling in probabilistic forecasting for irregular multivariate time series, where coupling between marginal distributions and dependency structures often introduces bias. To resolve this, we propose CopFITi, a novel model that decouples marginals from dependencies: it employs normalizing flows to flexibly model univariate marginal distributions and leverages a Gaussian mixture copula to capture complex multivariate dependencies. CopFITi is the first copula-based model for irregular multivariate time series that is explicitly constructed to satisfy marginal consistency by design. Experimental results demonstrate that CopFITi achieves state-of-the-art performance in joint density estimation, significantly improving both marginal calibration and overall predictive accuracy.

This study addresses the lack of a systematic approach for modeling distorted copulas and stable tail dependence functions (stdfs) within the Morillas-type framework. The authors propose a stochastic representation and a general sampling algorithm for such transformations, establishing a distortion framework tailored to stdfs. By leveraging monomial distortions and convex combinations, the method substantially enhances modeling flexibility for extremal dependence structures. Notably, they prove that monomial distortions with exponent less than one preserve the stdf property—a result that enables, for the first time, effective sampling from Morillas-type copulas. Furthermore, the work introduces a new class of distortion functions that maintain the extremal copula structure, extends modeling capabilities to non-regularly varying settings, provides efficient sampling schemes for Archimedean and Archimax copulas, and clarifies the impact of distortions on limits of max-domains of attraction.