85 papers
This paper introduces a reproducible, hybrid SIRD framework that estimates time-varying epidemic parameters via algebraic inversion and Vector Autoregression to outperform static and naive forecasting baselines in predicting COVID-19 cases and deaths, though it currently exhibits limited probabilistic calibration.
This paper introduces a machine learning approach using an adapted Partial Logistic Artificial Neural Network (PLANN) to estimate and predict net survival, demonstrating its flexibility in handling complex data structures compared to traditional spline-based regressions while noting higher variance in estimates for smaller sample sizes and providing the `survivalPLANN` R package for implementation.
This study introduces a graphon-based framework to analyze the structural evolution of Hungary's higher education mobility networks from 2006 to 2024, demonstrating that spectral and distributional distance metrics outperform traditional statistics in capturing policy-induced shifts and revealing hierarchical network tiers.
This study proposes an integrated economic–statistical framework for monitoring multivariate categorical processes in Phase II by optimizing the design of an EWMA–GLT control chart to minimize expected total costs while maintaining required statistical performance.
This paper proposes Transport G-Computation (TGC), a novel distributional framework combining longitudinal g-computation with optimal transport to estimate Wasserstein causal effects, which demonstrates superior performance over parametric g-computation in scenarios involving confounded feedback and model misspecification despite higher computational costs.
This paper introduces a geometrically motivated -deformed Hamiltonian Monte Carlo framework that utilizes Jackson-type differences to construct a Metropolis-adjusted sampler, offering a robust alternative to classical HMC that avoids domain violations and scale mismatches while converging to standard dynamics as .
This paper introduces CALIB-SAE, a non-stationary Gaussian process correction to the Fay-Herriot estimator that maintains equivalent point-prediction accuracy while significantly improving calibrated predictive uncertainty for small area estimation across both simulated and real-world geographic datasets.
This paper proposes a two-component finite mixture model combining Fréchet and Weibull distributions to address heterogeneous tail behavior in fields like hydrology and reliability, detailing its maximum likelihood estimation via the EM algorithm, validating its performance through simulations, and demonstrating its practical utility on real wind speed data.
This paper establishes the exact and asymptotic distributional properties of maximum likelihood estimators for the Lorenz curve and Gini index under the geometric distribution, providing a rigorous inferential framework for inequality measures in discrete settings through closed-form derivations, consistency proofs, simulation studies, and real-data application.
This paper demonstrates that self-exciting networks driven by a shared nonstationary background can appear critical due to statistical aliasing of the external drive into the interaction matrix, a mechanism that causes subcritical systems to be misdiagnosed as critical unless the common drive is explicitly corrected for.