longoR
It is the most common way of getting rid of nuisance parameters. Say, your parameters are of dimension p>1. But you are only interested in q dimension(s) of them. The the profile likelihood is to maximize the usual likelihood over the dimensions you are "not" interested in. Hence the resulting profile likelihood map from q dimension(s) to (0,1), instead of from p dimensions to (0,1).
orange
谢谢,好像明白了一点了!
我想把模型放上去让你帮我看看的,可是怎么都粘帖不上去啊,郁闷!
文章模型假设y=log(x+a) (1)
y服从正态分布,a为门限参数。
将a的极大似然估计值代入公式(1)
得到profile likelihood function,再去估计其他参数
(很想把公式粘帖上去让你帮我解释一下,可是弄不上去,不知道你能看明白我的意思不,如果能看懂,能不能再给我说具体点,不胜感激!)