A Bayesian zero-augmented unit-Weibull quantile mixed model with application to COVID-19 mortality in Asia

Abstract Mixed-effects models are widely used to examine associations between correlated outcomes and set of covariates. However, in real-world outcome measurements are continuous and bounded within the interval of 0 and 1, exhibit correlation, skewness, and contain zeros; thus, existing conventional mixed-effects models are not suitable. Therefore, a Bayesian zero-augmented unit-Weibull quantile mixed-effects model that accommodates these complex data structures is proposed. Following the model development, Monte Carlo simulations were used to evaluate the model performance, and the results confirmed that the all true parameter values were recovered effectively. The model was applied to examine association between COVID-19 death rates and country-level factors, including vaccination rate, proportion of population aged $$\ge 65$$ years, hospital bed count, population density, income level, and interaction between GDP and income level in Asian countries, with temporal trends captured using a piecewise linear time specification. The Bayesian approach was used for parameter estimation, with the No-U-Turn Sampler (NUTS), a variant of Hamiltonian Monte Carlo (HMC), implemented via the rstan package in R. The results from the continuous bounded component, which models the conditional median show that COVID-19 deaths increased until month 10, declined between months 10 and 14, rose again from months 14 to 20, and then declined thereafter. Furthermore, within-country vaccination, between-country vaccination, hospital bed count, and the GDP–income interaction were negatively associated with COVID-19 death rates, whereas population density and the proportion of the population aged $$\ge 65$$ years were positively associated with COVID-19 death rates. Likewise, the zero-augmented component show that between-country vaccination rate and hospital bed count increase the probability of observing zero COVID-19 death rates, whereas within-country vaccination rate and population density decrease the probability of observing zero deaths. The positive GDP–income interaction shows that the negative association between GDP and the probability of observing zero COVID-19 deaths is weaker in the upper-income group than in the lower-income group. Finally, the model comparison analysis highlighted that the proposed zero-augmented unit-Weibull quantile mixed model provided a better fit to the COVID-19 death rate data than the zero-augmented beta mixed model. The methodological development presented in this paper provides a statistical model for analyzing zero-augmented, bounded, and skewed response variables with a set of covariates. Its application to COVID-19 death rates in Asia highlights key factors associated with variations in death rates across countries, providing useful evidence for public health planning.

Authors

Institutions

Publication Details

Journal
Health Services and Outcomes Research Methodology
Published
2026-11-16
DOI
https://doi.org/10.1007/s10742-026-00400-0
Primary Topic
Statistical Distribution Estimation and Applications
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

A Bayesian zero-augmented unit-Weibull quantile mixed model with application to COVID-19 mortality in Asia

Nirajan Bam
2 links
Health Services and Outcomes Research Methodology
Statistical Distribution Estimation and Applications
article

A Bayesian zero-augmented unit-Weibull quantile mixed model with application to COVID-19 mortality in Asia

Nirajan Bam
article en

Abstract

Abstract Mixed-effects models are widely used to examine associations between correlated outcomes and set of covariates. However, in real-world outcome measurements are continuous and bounded within the interval of 0 and 1, exhibit correlation, skewness, and contain zeros; thus, existing conventional mixed-effects models are not suitable. Therefore, a Bayesian zero-augmented unit-Weibull quantile mixed-effects model that accommodates these complex data structures is proposed. Following the model development, Monte Carlo simulations were used to evaluate the model performance, and the results confirmed that the all true parameter values were recovered effectively. The model was applied to examine association between COVID-19 death rates and country-level factors, including vaccination rate, proportion of population aged $$\ge 65$$ years, hospital bed count, population density, income level, and interaction between GDP and income level in Asian countries, with temporal trends captured using a piecewise linear time specification. The Bayesian approach was used for parameter estimation, with the No-U-Turn Sampler (NUTS), a variant of Hamiltonian Monte Carlo (HMC), implemented via the rstan package in R. The results from the continuous bounded component, which models the conditional median show that COVID-19 deaths increased until month 10, declined between months 10 and 14, rose again from months 14 to 20, and then declined thereafter. Furthermore, within-country vaccination, between-country vaccination, hospital bed count, and the GDP–income interaction were negatively associated with COVID-19 death rates, whereas population density and the proportion of the population aged $$\ge 65$$ years were positively associated with COVID-19 death rates. Likewise, the zero-augmented component show that between-country vaccination rate and hospital bed count increase the probability of observing zero COVID-19 death rates, whereas within-country vaccination rate and population density decrease the probability of observing zero deaths. The positive GDP–income interaction shows that the negative association between GDP and the probability of observing zero COVID-19 deaths is weaker in the upper-income group than in the lower-income group. Finally, the model comparison analysis highlighted that the proposed zero-augmented unit-Weibull quantile mixed model provided a better fit to the COVID-19 death rate data than the zero-augmented beta mixed model. The methodological development presented in this paper provides a statistical model for analyzing zero-augmented, bounded, and skewed response variables with a set of covariates. Its application to COVID-19 death rates in Asia highlights key factors associated with variations in death rates across countries, providing useful evidence for public health planning.

Health Services and Outcomes Research Methodology
Miami University (US)
Good health and well-being
Statistical Distribution Estimation and Applications
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon