增值评价中缺失数据对线性回归模型成绩预测的影响

The Impact of Missing Data in Value-added Evaluation on the Performance Prediction of Linear Regression Models

在探索适合区域学校效能的增值评价方法时,线性回归模型因其具有较高的预测力与易理解的特点,成为入门的首选方法之一。在实际应用中,由于学生转学、回生源地参加中、高考等因素,建模时并不能获得全区域的完整数据,学校在计算效能时也受到一定的影响。本研究通过数据模拟的方式,使用线性模型,探讨了数据缺失对学校增值评价的影响。研究发现:(1)缺失比例较小(5%、10%)时,对学校增值效能计算的影响较小;(2)样本量对学校增值效能计算的影响较小;(3)当学校水平存在差异(好、中、差)时,缺失数据引起的增值效能计算误差大于学校水平均衡时的误差。

When exploring the value-added evaluation method suitable for regional school effectiveness,the linear regression model is one of the preferred methods for entry because of its high predictive power and easy understanding.In practical applications,due to factors such as student transfer and returning to the ground source for senior high or college school entrance examination,the complete data of the whole region cannot be obtained during modeling,and the school is also affected when calculating effectiveness.This study explores the impact of missing data on school value-added evaluation by means of data simulation and linear models.The study found that:(1)when the proportion of missing data is small(5%,10%),the impact on the school's value-added effectiveness calculation is small;(2)the sample size has less impact on the school's value-added performance calculation;(3)when there are differences in school levels(good,medium,and poor),the calculation error of valueadded effectiveness caused by missing data is greater than the error of school level equilibrium.