一种简单有效的二项分布比例参数似然比置信区间的求法

对二项分布比例参数p的似然比置信区间,提出一种简便求解方法。在平均覆盖率、平均区间长度及区间长度的95%置信区间准则下与WScore、Plus4、Jeffreys置信区间进行模拟比较。试验表明,在二项分布b(n,p)的参数n≥20且p∈(0.1,0.9)时,该方法获取的似然比置信区间性能优良。当点估计p值不是接近于0或1且n≥20时,推荐使用本方法获取p的置信区间。

A Simple and Efficient Solution of Likelihood Ratio-based Confidence Interval for the Binomial Proportion

A simple and efficient solution method for likelihood ratio-based confidence interval for the binomial distribution proportion parameter was proposed, and in terms of "average coverage", "average interval length", "the 95% confidence interval of a interval length" criteria, the performance of the likelihood ratio-based confidence interval obtained by our method was compared with WScore, Plus4, Jeffreys confidence intervals by Monte Carlo simulation.Simulation results demonstrate that in case of n≥20 and p is not small, or not large, the likelihood ratio-based confidence interval obtained by our method has excellent performance.Based on the results, when the point estimation value of the binomial distribution proportion parameter is not approach to 0, or to 1, we recommend using the likelihood ratio-based confidence interval obtained by our method.