Extreme climate impacts are often managed at neighborhood and local administrative scales, yet reanalysis products remain spatially coarse and are not aligned with administrative boundaries such as Census Block Groups (CBGs). This study develops a machine-learning-based change-of-support reconstruction workflow to translate hourly ERA5-Land 2-m air temperature (2000–2024) from its native approximately 9-km grid to CBG areal support across Maryland. ERA5-Land fields were mapped to CBGs using sparse bilinear weights and decomposed into a leave-one-year-out hourly climatology and CBG-specific anomalies. Month-specific Random Forest models reconstructed ERA5-Land-derived anomalies using temporal lags, harmonic terms, dew point, 10-m winds, and static spatial predictors. Validation against 14 independent daily station records showed annual root-mean-square error (RMSE) values of 1.2–3.1 °C and Pearson correlations of 0.95–0.99, with larger errors in winter and high-elevation CBGs. A supplementary 2024 subdaily consistency check using four independent NOAA/NCEI airport stations and 34,471 common station-hours yielded a pooled hourly RMSE of 1.86 °C and correlation of 0.983, with larger errors during nighttime than daytime hours. Before bias adjustment, relative to deterministic ERA5-to-CBG mapping using nearest-neighbor and inverse-distance weighting, the reconstructed series had lower RMSE at 11/14 and 12/14 stations, respectively; after leave-one-year-out bias adjustment, the corresponding counts were 10/14 and 9/14. In unadjusted seasonal comparisons, RMSE reductions reached 14% relative to nearest-neighbor mapping and 6.84% relative to inverse-distance weighting, although overall gains were modest. Targeted sensitivity analyses showed that temporal-memory predictors dominated model importance, whereas contemporary landscape variables and direct geographical coordinates contributed comparatively little. The resulting hourly CBG-support temperature series provides administrative-unit climate information for ecological, heat-risk, public-health, and climate-adaptation research; applications that depend strongly on local hourly extremes should include additional application-specific validation.
Assessing the statistical dependence between a nominal response variable and a collection of explanatory variables remains a challenging problem in multivariate analysis. Consequently, there is a need for multiple association measures that are readily interpretable and suitable for the analysis of categorical regression models. This paper proposes a general framework for constructing multiple association measures between a target variable and a vector of explanatory variables. The proposed framework is complemented by sample counterparts for point estimation and by a comprehensive inferential methodology, including confidence interval construction and hypothesis testing based on bootstrap resampling and permutation procedures. Building upon this framework, three new multiple association measures are developed for a nominal response variable and a set of explanatory variables with mixed measurement scales. The proposed measures are derived from a generic categorical regression model and well-established bivariate association measures for nominal variables, namely Goodman–Kruskal’s lambda and tau, and Cohen’s kappa coefficient. They are computationally straightforward, readily interpretable, and broadly applicable across a wide range of scientific disciplines. The performance of the proposed measures and their associated inferential procedures is evaluated through simulation studies covering different levels of statistical dependence. Finally, the methodology is illustrated by analyzing the multiple association between consumer confidence levels and a set of demographic characteristics and opinion variables using data from a social survey.
Network psychometrics has emerged as a powerful framework for understanding cognitive functioning, yet clustering individuals based on network dissimilarity poses methodological challenges. Traditional approaches that apply statistical thresholding during cluster optimization risk circular analysis and overfitting. We propose a two-phase network-based clustering method that separates exploratory cluster identification from confirmatory inference. Phase 1 optimizes cluster assignments using weighted, unthresholded correlation networks and comparing results across three dissimilarity metrics (Frobenius distance, DeltaCon, and spectral distance). Phase 2 characterizes identified clusters using permutation-based network estimation with statistical thresholding. The proposed approach was validated on simulated data and applied to a motivating empirical cognitive dataset, where it was compared with traditional single-phase methods and evaluated in terms of stability across dissimilarity metrics. In simulations with known ground truth, Frobenius distance provided the best partition recovery across all scenarios, outperforming DeltaCon and spectral distance. Although individual-level recovery was moderate, the Frobenius-based algorithm successfully reproduced the main population-level network contrasts, particularly in signed and sparse-versus-dense structures. Applied to cognitive data from N=4471 participants across 7 cognitive domains, the two-phase approach avoided inflated dissimilarity estimates while maintaining computational efficiency. The final Frobenius-based empirical solution identified two qualitatively distinct cognitive network profiles: a sparser, more modular profile characterized by domain-specific associations, and a dense, globally integrated profile with stronger connectivity across cognitive domains. The two-phase framework provides methodologically rigorous network-based clustering that avoids circular analysis. This approach is applicable to psychological, neuroscience, and clinical research where network structure differs across subgroups.
Estimating the number of events in a time frame and/or area is a central problem in contexts of armed conflict and humanitarian crises. Documentation of victims is invariably lacking, with substantial selection bias and undercount. Multiple-systems estimation methods address this problem by leveraging overlap across multiple sources. We provide a selective overview of population size estimation in humanitarian upheavals, focusing on the interplay between data collection and statistical modeling. We discuss key challenges in data construction and integration, and review major classes of approaches. Methods are illustrated using datasets from recent conflict and migration contexts.
We present a new class of survival models that combine unobserved population heterogeneity and cured individuals by incorporating frailty terms from the Power Variance Function (PVF) family, referred to as PVF-Induced Defective (PVF-ID) models. This structure broadens classic frailty theory and defective lifetimes under a proportional-hazards lens, letting researchers flexibly track long-term survivors in mixed cohorts. Inference follows a full-Bayesian routine run via Hamiltonian Monte Carlo, and key frequentist properties are cross-checked using simulation studies. When applied to HIV/AIDS records, the model reliably pinpoints important covariates and quantifies the fraction of cured patients. In this case study, among several baseline hazards, the simple exponential form wins by both leave-plus-one marginal likelihood (LPML) and raw marginal likelihood. Together, these findings position the PVF-ID framework as a robust, versatile toolbox for studying survival data with immune subpopulations.
