作者xx801114 (狮子)
看板Statistics
标题[程式] 费雪信息矩阵计算
时间Sat May 16 23:07:38 2020
[软体程式类别]:R
[程式问题]:
计算费雪信息矩阵
[软体熟悉度]:
新手
[问题叙述]:
可靠度的资料,采用Lognormal来计算
资料分别有Exact failure time跟censored .
把likelihood function 打出来後,想取Log後利用微分来计算费雪信息数,但是R好像没
办法直接对lognormal的pdf及cdf微分,想请问各位大大,有什麽方式可以计算费雪信息
矩阵? 谢谢大家
[程式范例]:
写出来的likelihood function:
likelihood=function(x,y){
dlnorm(6700,x,y)*(1-plnorm(6950,x,y))*(1-plnorm(7820,x,y))*(1-plnorm(8790,x,
y))*dlnorm(9120,x,y)*(1-plnorm(9660,x,y))*(1-plnorm(9820,x,y))*
(1-plnorm(11310,x,y))*(1-plnorm(11690,x,y))*(1-plnorm(11850,x,y))*(1-plnor
m(11880,x,y))*(1-plnorm(12140,x,y))*dlnorm(12200,x,y)*
(1-plnorm(12870,x,y))*dlnorm(13150,x,y)*(1-plnorm(13330,x,y))*(1-plnorm(13
470,x,y))*(1-plnorm(14040,x,y))*dlnorm(14300,x,y)*
dlnorm(17520,x,y)*(1-plnorm(17540,x,y))*(1-plnorm(17890,x,y))*(1-plnorm(18
450,x,y))*(1-plnorm(18960,x,y))*(1-plnorm(18980,x,y))*
(1-plnorm(19410,x,y))*dlnorm(20100,x,y)*(1-plnorm(20100,x,y))*(1-plnorm(20
150,x,y))*(1-plnorm(20320,x,y))*dlnorm(20900,x,y)*dlnorm(22700,x,y)*
(1-plnorm(23490,x,y))*dlnorm(26510,x,y)*(1-plnorm(27410,x,y))*dlnorm(27490
,x,y)*(1-plnorm(27890,x,y))*(1-plnorm(28100,x,y))
}
--
※ 发信站: 批踢踢实业坊(ptt.cc), 来自: 39.9.105.183 (台湾)
※ 文章网址: https://webptt.com/cn.aspx?n=bbs/Statistics/M.1589641660.A.36E.html
1F:推 ksherry: 可以考虑 numerical Hessian matrix 05/17 20:21
2F:→ xx801114: 收到,谢谢您 05/18 09:28
3F:推 cuylerLin: 对数常态的信息矩阵理论上推导得出来啊,有closed form 05/18 23:32