作　　者： ; ; ; ; ;

机构地区： 华南师范大学

出　　处： 《中国考试》 2009年第1期9-15,共7页

摘　　要： 采用数据模拟方法分析分位数回归（QR）和最小二乘回归（OLSR）之间的异同。结果表明：QR可以克服OLSR强假设、易受极端值影响、只能描述总体的平均信息等不足，充分揭示条件分布函数各点的局部信息，提高考试研究的针对性。尤其在随机误差项方差不恒定时，分位数回归的优势较最小二乘回归更为明显。它能挖掘考试数据中更多更有用的信息。 The differences between Quantile Regression （QR） and Ordinary Least-Square Regression（OISR） are analyzed by simulation. The results show that QR could overcome many of shortages of OLSR, such as requiring strong assumption, vulnerable to extreme＇s interference, describing the overall average and so forth. QR fully uncovers all parts of the local information of the conditional distribution function and enhances the efficiency in examination study. Especially, the advantages of QR are more obvious if the variance of residuals is not constant, as it could mining more plentiful and useful information in the data of examinations.

关 键 词： 心理统计 最小二乘回归 分位数回归

领　　域： [文化科学] [文化科学] [文化科学]