核等值:一种观察分数等值体系

Kernel equating: A framework of observed score equating

核等值流程包括:预平滑、估计分数概率、连续化、等值、评估等值结果。该方法兼具线性等值与等百分位等值的优点,各环节扩展性与包容性较强;采用平滑与连续化处理,可降低等值随机误差;等值差异标准误等其所特有的概念为结果评估提供可靠的工具。连续化与带宽选择方法等因素均可影响其表现;基于核等值的新方法为等值发展提供了新颖的视角。未来可关注核等值体系的扩充与完善、流程的更新、等值方法的结合和比较等方向。

Kernel equating procedures include pre-smoothing, estimation of the score probabilities, continuization, equating, and evaluation of equating performance. By incorporating linear equating and equipercentile equating methods, kernel equating is more extensible and comprehensive. Pre-smoothing and continuization are distinctive features in kernel equating to reduce the standard error of equating. Standard error of the difference between equating functions are calculated as criterion for evaluating the performances of different kernel equatings. Continuization methods, bandwidth selection methods, etc., can affect the performance of kernel equating. New equating methods based on kernel equating provide an innovative perspective for researchers. Further researchers could focus on extending kernel equating framework by integrating other methods, updating smoothing procedures, and comparative studies.