有序响应变量的贝叶斯模型选择及其在COPD疾病防治中的应用

慢性阻塞性肺病(COPD)是一种发病率、死亡率都非常高的疾病,且COPD的诊断和严重程度分级依赖于肺功能的检查,但是由于肺功能检查仪器价格昂贵,使得这项检查在很多经济欠发达地区尤其是农村基层医院并没有普及。本文基于有序响应变量模型致力于研究一种便于基层和社区使用的可以初步判别COPD病情的模型,以期提高我国基层和社区的COPD 防治水平。利用贝叶斯变量选择方法和数据增强的潜变量策略得到了易于实施的Gibbs后验抽样算法。数值模拟分析进一步说明了本文提出的有序响应变量贝叶斯模型选择方法的有效性,实例分析得到了易于判别COPD严重程度的稀疏模型。

Bayesian Model Selection of Ordinal Response Variables and Its Application in Disease Prevention of COPD

Chronic Obstructive Pulmonary Disease( COPD) is a disease with high morbidity and mortality.The diagnosis and severity grading of COPD depend on the examination of lung functions. However,due to the high cost of lung function examination instruments,this examination is not widely used in many economically underdeveloped areas, especially in rural grassroots hospitals. This paper studies how grassroots and communities can preliminarily identify COPD conditions with ordinal response model,so as to improve the level of prevention and treatment of COPD in grassroots and communities in China. Based on the Bayesian variable selection method and the data augmentation approach of latent variable,we obtain the Gibbs posterior sampling algorithm. Numerical simulation analysis further illustrates the usefulness of the proposed Bayesian model selection method for ordinal response variables,and the sparse models are given to determine the severity of COPD for real data analysis.