论文部分内容阅读
[目的]阐明定群监测计数资料零截尾分布数据的分析方法。[方法]模拟研究零截尾计数分布数据的参数估计;应用呼吸机使用时间的实例进一步完成截尾计数模型的实现。[结果]模拟研究表明零截尾Poisson模型参数估计值与真实参数值很接近,而传统的Poisson回归估计值与真值相差较大;呼吸机使用时间过度离散,似然比检验拒绝零截尾Poisson模型,采用截尾负二项模型更合适。实例分析表明有并发症、气管切开、APACHE得分越高、血中性粒细胞比例增高、伴有肺部炎症等是影响患者呼吸机使用天数增多的主要因素。[结论]截尾计数模型可以很好地解决定群监测零截尾分布计数资料的分析问题。
[Objective] The research aimed to clarify the analysis method of zero-censored distribution data of population monitoring and counting data. [Method] The parameter estimation of zero-censored counting distribution data was simulated and the realization of censored counting model was further implemented by using the example of ventilator usage time. [Results] Simulation results show that the estimated parameters of zero-truncated Poisson model are close to the true values, while the traditional estimates of Poisson regression differ greatly from the true values. Excessive ventilator time is used, and the likelihood ratio test rejects the zero- Poisson model, the use of truncated negative binomial model is more appropriate. Case analysis showed that there are complications, tracheotomy, APACHE score higher blood neutrophil increased, accompanied by pulmonary inflammation are the main factors affecting the patient’s use of ventilator days increased. [Conclusion] The censored count model can well solve the analysis of censored zero count distribution data.