※ 引述《bluejaykal (黑嘛嘛)》之铭言： : 想以模拟的方式比较 对於多重复发时间的情况之下 : margianl model 以及failty model 对於共变数的估计 : 所以以classical linear regression approach的方式 : ie. X=logY=u+beta*z+sigma : sigma服从extreme value distribution : 生成了一组服从独立韦伯分布的时间数据(X) : 但是多重复发时间常含有相关性存在 : 想将相关性加入 : 但不知有没有办法能从 : 已生成的独立韦伯分布时间 : 加入相关性 : 想请问各位大大 有没有什麽办法 : 感激不尽 "多重复发时间"指的是recurrent event data的setting吗? If yes, are you interested in modeling the recurrent events of a subject as a renewal process or as a counting process? The former focuses on the distribution of gap times (the time between two consecutive events), and the latter focuses on intensity or rate function. The two types of approaches require different estimation methods, and usually they are not comparable with each other. I assume you are interested in modeling and estimation of gap time distribution, judging from your original question. If this is the case, you should read the simulation section in the following papers Wang and Chang (JASA 1999) Huang (JRSSB 2002) Huang and Chen (LIDA 2003) Each of them simulated recurrent event data in a different way. ※ 编辑: bmka 来自: 71.251.54.131 (04/17 20:02)