Mixed Semi-Continuous Copula Additive Regression Models with Applications in Insurance and Health Care

保险和医疗保健数据通常呈现高比例的零，可能是偏斜的正值。例如，保险索赔和医疗支出属于那些持续扭曲的数据，峰值为零。此外，这些数据是相互关联的，研究人员可能希望将它们联合起来建模。例如，可以同时对不同类型的医疗支出（住院和门诊）、健康结果（质量调整生命年）和成本以及同一保险单提供的多种保险类型进行建模，同时考虑到它们的依赖性及其分布的特殊性（高比例的零与正值）。考虑到这两个特征确实很重要，原因包括获得无偏估计、可靠的置信区间、提供可靠的预测等等。我们的目标是提出一种强大、灵活和复杂的方法，能够以有效和高效的方式处理这些数据的性质，同时产生易于理解的结果，特别是风险分析人员和卫生决策者易于解释和理解。

Insurance and health care data often present a high proportion of zeros with possible skewed positive values. For example, insurance claims and medical expenditures are among those continuous skewed data with a spike at zero. Furthermore, these data are interconnected and researcher might wish to model them jointly. For instance, the interest can be in modelling simultaneously different types of medical expenditure (inpatient and outpatient), health outcomes (quality adjusted life years, QALYs) and costs, and multiple types of coverage provided by the same insurance policy, while accounting for both their dependence and the particular nature of their distributions (high proportion of zeros with positive values). Accounting for both features is indeed critical for many reasons including obtaining unbiased estimates, reliable confidence intervals, providing reliable predictions, to name but a few. Our aim is to propose a powerful, flexible and sophisticated approach that can deal with the nature of these data in an effective and efficient way while producing comprehensible results that are easy to interpret and understand, particularly by risk analysts and health decision makers.