本文主要讨论预测样本变化对题目难易度产生的影响,这种影响直接关系到在不同样本中预测的不同题目之间如何进行混合拼卷的问题。我们的解决方法是运用一元线性回归方程进行难易度参数的转换,使不同预测样本获得的题目参数可以统一到一个尺度上拼正式卷。 This article mainly discusses the impact of predicting sample changes on the degree of difficulty of a topic. This kind of impact is directly related to the problem of how to combine and uncoil the different topics predicted in different samples. Our solution is to use unary linear regression equations to convert the difficulty parameters so that the topic parameters obtained by different prediction samples can be unified into a standard scale.