中文摘要：

偏态抗差箱形图的构建函数决定了方法对偏态异常数据的探测风险和探测效率．目的是要分析选择水文系统中降雨异常值探测的偏态抗差箱形图的构建函数．论文选用了线性函数、二次曲面函数和幂函数3种函数形式，分别作用于无异常值数据和有异常值数据．结果显示，线性函数与二次曲面函数对于箱形图下边缘的拟合误差较大，仅幂函数对于上下边缘的拟合效果较好；幂函数对于无异常值数据的探测风险最小，接近于标准箱形图作用于对称正态分布数据效果；对于异常数据，幂函数的探测效果最接近于异常值污染比例，说明抗差性能强．幂函数最终确定为所选函数．幂函数构建的偏态抗差箱形图的上下边缘区间向偏度方向转移，使其比标准箱形图更加适应于偏态数据．

英文摘要：

The function to build the skewed robust boxplot that is the key to detecting risk and efficiency for skewed outlying data. This function is applicable to skewed outlying precipitation data in hydrological system which is focus. Uncontaminated data and contaminated data are analyzed by means of the linear function and quadratic function and exponential function. Research results indicate that.. 1） There is great error in fitted regression curves for the lower whisker by linear and quadratic functions. However, the ex- ponential function is appropriate for both the lower whisker and upper tail. 2） The detecting risk of the ex- ponential function is least at uncontaminated data, which substantially equaled that of standard boxplot at a symmetric normal distribution. 3） The detecting efficiency of the exponential function is closer to the percentage of contamination than that of the other two functions. The exponential function is the best one for skewed robust boxplot; and its lower whisker and upper whisker shifted to the skewness which is more applicable to the skewed data than standard boxplot.