家畜生产中生存模型的分析方法及比较研究

Comparative study on analysis methods of the survival models in livestock production

本文介绍了生存分析的概念和特点。对线性模型、随机回归模型和比例风险模型等多种生存模型在家畜生产中的应用进行了比较,重点讨论了比例风险模型的分布、非参数估计以及该模型的扩展应用。结果表明:线性模型适合把生存性状定义成二元性状进行分析,但局限是很难解释截尾问题;随机回归模型可作为利用线性模型来分析截尾数据,但当数据量较少或缺失时,很难处理时间依赖型变量;比例风险模型尽管目前尚不能在超过两个性状的多元分析中得到应用,但因其比较适合处理生存数据,特别是分析截尾数据或时间依赖型变量,在动物育种数据分析时有广泛应用。

This paper introduced the concept and characteristics of survival analysis. The linear model, random regression model and theproportional hazard model and other survival models in domestic animals were compared. The distribution, non-parametric estimation and theexpansion of the model application about the proportional hazard model were discussed. The results showed that the definition of survival traitsas binary traits in the linear model were analyzed, but the limitation was difficult to explain the truncation problem; random regression modelcould be used as a linear model to analyze censored data, but when the data quantity was less or missing, it was difficult to deal with time de-pendent variables; although proportional hazard model was not used in the multivariate analysis of more than two traits, but because of theconvenience for survival data, especially the analysis of censored data or time dependent variables, there was extensive application in animalbreeding.