|
|
本帖最后由 mono 于 2014-6-6 12:33 编辑
95%可信区间是用正态分布来估计的，你问问你自己，为什么你的95%的可信区间总是对称的，你可以用profile 95%CI 来坐估计，这样你的95%的可信区间就不是对称的，卡放分布本身就不是对称的。
读读下面的东西，可能对你有益。
In general, the confidence interval based on the standard error strongly depends on the assumption of normality for the estimator. The "profile likelihood confidence interval" provides an alternative.
I am pretty sure you can find documentation for this. For instance, here and references therein.
Here is a brief overview.
Let us say the data depend upon two (vectors of) parameters, θ and δ , where θ is of interest and δ is a nuisance parameter.
The profile likelihood of θ is defined by
L p (θ)=max δ L(θ,δ)
where L(θ,δ) is the 'complete likelihood'. L p (θ) does no longer depend on δ since it has been profiled out.
Let a null hypothesis be H 0 :θ=θ 0 and the likelihood ratio statistic be
LR=2(logL p (θ ^ )−logL p (θ 0 ))
where θ ^ is the value of θ that maximises the profile likelihood L p (θ) .
A "profile likelihood confidence interval" for θ consists of those values θ 0 for which the test is not significant.
|
|
发表于 2014-6-5 11:54:21
微信
收藏
楼主
