负二项抽样因其在发病率很低的情况下的优良表现而被广泛应用于流行病学及其它学科之中. ``需处理数''是一种度量药物疗效的重要指标, 它常常用来评价那些结果是二值变量的临床试验所 研究的药物的疗效. 在实际应用中, 通常希望得到需处理数的置信区间, 但是目前已有的需处 理数的置信区间构造方法都存在一个应用上的难题: 区间上限过大以至于不可靠. 文章旨在解决需 处理数区间上限估计过大的问题, 为此提出了需处理数的最短区间构造方法并运用蒙特卡洛模 拟方法比较其相对传统方法的优劣, 还给出了实际应用的例子. 模拟结果表明: 改进后的方法能够 在控制置信系数的情况下极大地减小区间上限,具有重要的实际价值.

Negative binomial sampling are wildy used in epidemiology and many other fields owing to its good performance in binary results clinical trails when the prevalence of the disease is rare. As a measurement of drug effect of randomized controlled trials with binary outcomes, the number needed to treat (NNT) is a useful way of reporting trials results. In clinical appliction, we prefer to report the confidence interval of the number needed to treat. The most popular confidence interval for a number needed to treat is the Wald type interval. Unfortunately, the upper confidence limit of Wald type interval often trends to be unreliable. In this paper, the shortest interval is proposed as an improved confidence interval for the number need to be treat which can reduce the upper confidence limit of the interval. Monte Calro Simulation method is used to compare the perfromance of the improved interval with the Wald type interval, and two illustratvie examples show that the improved interval has a coverage probablity close to the confidence coefficient 95% and can reduce the upper  onfidence limit of interval significantly, which is of practical importance.