Tail-Sensitive Modeling of Medical Insurance Costs: A Covariate-Aware Analysis
Qubahan Educational Journal · 16.09.2026 · 3 (3) · 1-15
DOI: 10.48161/qej.v3n3a19 ↗Kurz gefasst
Fachartikel in der Zeitschrift Qubahan Educational Journal, online erschienen am 16.09.2026. Verfasst von Abeda Muhammad Iqbal und 2 weiteren Autor:innen. Die Arbeit nennt 30 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY-NC-ND 4.0).
Abstract (Originalsprache)
Selecting an appropriate probability distribution is fundamental to reliable statistical modeling because incorrect distributional assumptions can distort estimation, goodness-of-fit interpretation, and applied decision-making. This study develops and empirically evaluates a covariate-sensitive distribution-fitting framework for medical insurance cost modeling using one real-world dataset. We selected the Medical Cost Personal dataset because it contains a clearly defined continuous positive outcome, medical charges, along with interpretable explanatory variables, including age, body mass index (BMI), number of children, smoking status, sex, and residential region. The analysis uses 1,338 observations and seven variables after confirming that the dataset contains no missing values. We use maximum likelihood estimation (MLE) to fit normal, lognormal, gamma, Weibull, and exponential distributions to the medical-charge outcome. We assess model adequacy using log-likelihood, Akaike Information Criterion (AIC), Bayesian Information Criterion (BIC), Kolmogorov–Smirnov (KS) diagnostics, Cramér–von Mises (CvM) diagnostics, Anderson–Darling (AD) statistics, probability plots, and empirical-versus-theoretical density comparisons. The analysis further links distributional behavior to covariate structure by examining subgroup patterns and estimating log-linear and gamma generalized linear models (GLMs). The findings show that medical charges are strongly right-skewed, with a mean of 13,270.42, a median below the mean, and a skewness value of 1.52. The lognormal distribution provides the strongest overall fit, producing the lowest Akaike Information Criterion (AIC) and Bayesian Information Criterion (BIC) values and the most acceptable visual agreement with the empirical distribution. Covariate-based results show that smoking status, age, body mass index (BMI), and number of children are associated with systematic differences in medical-charge levels and distributional shape. The study offers a practical framework for selecting probability distributions in applied cost modeling and shows that distribution choice should be guided by outcome scale, skewness, tail behavior, and covariate-linked heterogeneity rather than mechanical model application.
Zitieren
APA
Iqbal, A. M., Narayanapillai, P. T., & Sivanesan, S. (2026). Tail-Sensitive Modeling of Medical Insurance Costs: A Covariate-Aware Analysis. Qubahan Educational Journal, 3(3), 1-15. https://doi.org/10.48161/qej.v3n3a19
Vancouver
Iqbal AM, Narayanapillai PT, Sivanesan S. Tail-Sensitive Modeling of Medical Insurance Costs: A Covariate-Aware Analysis. Qubahan Educational Journal. 2026;3(3):1-15. doi:10.48161/qej.v3n3a19
BibTeX
@article{Iqbal2026,
author = {Iqbal, Abeda Muhammad and Narayanapillai, Padmesh Tirunelveli and Sivanesan, Sivakumar},
title = {Tail-Sensitive Modeling of Medical Insurance Costs: A Covariate-Aware Analysis},
journal = {Qubahan Educational Journal},
year = {2026},
volume = {3},
number = {3},
pages = {1--15},
doi = {10.48161/qej.v3n3a19}
}
EUVIMED