摘要
首先,证明了基于Bootstrap法的BCa区间二阶矫正性质的相关抽象结论;其次,讨论了BCa区间的算法实现,并进行了实证分析。结果表明,在分布左尾较长的情形下,分位数法将正态区间的端点向左调整,BCa法又将分位数区间端点进一步向左调整,区间对称性也从正态区间的1,调整至分位数区间的0.58和BCa区间的0.38。这些调整均说明BCa区间是最为符合分布特性的估计区间,印证了BCa区间的优良二阶性质。
Firstly,abstract conclusions of second- order correct property of BCa interval based on bootstrap method are simplified. Then,BCa interval algorithm is discussed,and example is used to show how this is done. The results show,in the case of long left tail distribution,standard interval endpoint is adjusted to the left by percentile method,and percentile interval endpoint is further adjusted to the left by BCa method; The shape about symmetry of standard interval is adjusted from 1 to percentile interval's 0. 58,and further to BCa interval's 0. 38. These adjustments show that BCa estimation interval is the most consistent with distribution characteristics,which also confirm excellent second- order properties of BCa interval.
出处
《延安大学学报(自然科学版)》
2014年第4期8-11,共4页
Journal of Yan'an University:Natural Science Edition
基金
国家自然基金(51379173)
天水师范学院中青年教师科研资助项目(TSY201204)