英语翻译刚才有位朋友回答的不错!已经被评为最佳答案了,文章很长,我会一点一点都贴出来.还是那句话：怎么翻译都好,用软件也行.不过要翻译的通畅（用软件翻译完了帮我改下不通的地方

Question

英语翻译
刚才有位朋友回答的不错!已经被评为最佳答案了,文章很长,我会一点一点都贴出来.还是那句话：怎么翻译都好,用软件也行.不过要翻译的通畅（用软件翻译完了帮我改下不通的地方）.数学符号很多,麻烦点,不太清楚的地方可以看原文哈.
3 CONFIDENCE INTERVAL P VALUE
Berger and Boos (1994) proposed a new method of computing a p value.In the problem of comparing two binomial proportions,if H is true,p = p1 = p2 is a nuisance parameter.Let ….denote a 100 ……confidence interval for p calculated from the data (x,y) and assuming p1 = p2 = p.The confidence interval used in this article is the Clopper and Pearson (1934) interval based on X + Y,a binomial (m + n,p) random variable if p1 = p2 = p.This interval is easily computed from the formula
…………………..
where a = x + y,b = m + n,and F is the upper 100 percentile of an distribution with v and n degrees of freedom.
The confidence interval p value,based on the statistic Z is defined by
………………………….
where R is the same as in the definition of pz.pc differs from pz in that the supremum is taken over the confidence interval C rather than over the whole range …..,and the error probability B is added to the supremum.If B = 0,pc is the same as pz.Berger and Boos (1994) showed that this modification of the usual definition of a p value yields a valid p value.That is,the test that rejects H if and only if pc (X,Y) < a is an unconditional level-a test.The error probability is specified by the experimenter.Different values of B yield different p values and tests.In this article,B=0.001 is used (as suggested by Berger and Boos).
Berger and Boos proposed the confidence interval based p value for two reasons.The First is computational.In both pz and pc,the function to be maximized is the same.The maximization over the smaller set,C,can be much simpler.The second is statistical.Having observed the data,we should be able to estimate p and should not need to consider values of p that are completely unsupported by the data.In pc,only those “plausible” values that are in the confidence interval are considered.
This article points out that the confidence interval p value can have a third advantage.It can produce tests with higher power that the usual p value.And,remember,this is achieved with less computational effort.

Answer 1

3 置信区间P值
Berger and Boos (1994)提出了一个新的计算P值的方法.在这个比较两个二项式比例的问题中,如果 是成立的,就是多余参量.设 表示一个从数据(x,y)中计算P 和假设 的 的置信区间.这篇文章中使用的置信区间是Clopper and Pearson (1934)以如果 ,那么X + Y,一个双项式 (m + n,p) 随机变量为基础的区间.这个区间很容易由公式
(3)
计算得出.
公式中 a = x + y,b = m + n,而 是一个带有 ν 和 η 自由度的F 分布的上部 100(β/2)百分比.
以统计量Z为基础的置信区间P值被表示为
在公式中Rz(x,y)与pz的定义相同.在上确界被带出了置信区间 而不是越出了全部极差 ,并且误差概率β被增加到了上确界时,pc不同于pz.如果β = 0,ps和pz相同.Berger and Boos (1994)表示这个对于P值的一般定义的修正产生了一个有效的P值.就是说,如果或仅仅如果 是一个无条件 level-α测试的话,那么测试否定 .实验者详细指明了误差概率.不同的β值得到不同的P值和测试.在这篇文章中,取β=0.001（按照Berger and Boos的建议）.
Berger and Boos有两个原因提议以P值为置信区间的基础.第一个是计算上的.在pz和pc两者中,最大化函数是相同的.越出较小集 的最大值化可以更为简化.第二个就是统计上的.通过观察数据,我们能够推断p而且不需顾及p的值完全无数据支持.在pc里,只有那些在置信区间中“看似正确”的值才会被考虑.
这篇文章指出置信区间P值能拥有第三个优势.它可以产生比一般P值更高推翻假设的机率的测试.还有,记住,这用较少的计算可以达到.