427 papers
This paper demonstrates that survey weighting on socio-demographic variables only removes a fraction of nonresponse bias equal to the proportion of outcome variance those variables explain, revealing that such weighting is largely ineffective for subjective well-being research where demographic predictors account for minimal variance in happiness.
This paper introduces the New Generalized Alpha Power (NGAP) family of continuous distributions, which utilizes an exponent transformation to enhance modeling flexibility for complex lifetime data, and validates its superior performance through comprehensive mathematical analysis, extensive simulation of fifteen estimation methods, and successful applications in healthcare and environmental quality monitoring.
This paper develops a likelihood-based theory demonstrating that in multiclass classification, informative label missingness can reduce asymptotic excess risk by aligning information gains from the missingness pattern with decision boundary perturbations, even when complete classification retains higher global Fisher information.
This paper proposes and mathematically validates a two-step statistical method for the International Chopin Competition that calibrates jurors' scoring scales and iteratively clips outliers to ensure convergence, thereby preserving artistic judgment while improving the fairness of score aggregation.
This paper demonstrates that machine learning-based feature selection can significantly reduce the computational and interpretive burden of multidimensional poverty measurement by identifying a small, non-redundant subset of indicators that maintains classification accuracy comparable to full indicator sets, as evidenced by an analysis of India's IHDS-II data.
This methodological study demonstrates that claims of a post-2022 decline in mean lexical diversity across the English web are not robust, as placebo tests reveal spurious breakpoints in multiple domains and the remaining statistically viable analyses are underpowered and inconclusive.
This paper presents a privacy-preserving computational framework that combines paired-area randomized experiments, secure aggregation, and differential privacy with hierarchical Bayesian inference to accurately estimate incremental local advertising effects and optimize budgets despite sparse conversion data and strict privacy constraints.
This paper corrects a sign error in the Shannon entropy formula for the Transmuted Weibull distribution and develops comprehensive frequentist and Bayesian inference methods for estimating its parameters and entropy under adaptive progressive Type-II censored samples, demonstrating through simulations and real data that the Bayesian approach generally outperforms maximum likelihood estimation, especially for small or heavily censored datasets.
This study employs a design-consistent, fairness-audited machine learning approach on PISA 2022 data from 143,573 disadvantaged adolescents across 79 education systems to identify that while mathematics self-efficacy is the strongest universal predictor of STEM resilience, predictive models exhibit significant transportability limitations and systematic biases against girls and rural students, underscoring the need for context-specific and equitable screening strategies.
This paper proposes a methodology that applies Gröbner basis theory to linear combination models in medical statistics to systematically identify and generalize confounding relations, thereby enhancing the accuracy of causal effect estimation in epidemiological studies.