It is known that a family of fixed-effects item response models with equal discrimination and different guessing parameters has no model identifiability. For this family, some types of information including the Fisher information and a new one are maximized to have model identification. The conditions of monotonicity of these types of information with respect to a tuning parameter are given. In the case of the logistic model with guessing parameters, it is shown that maxima do not exist under some parametrization, where negative lower asymptote can be employed without changing the probabilities of correct responses by examinees.

中文翻译：

---

具有猜测参数的未识别项目响应模型的某些类型信息的最大化

众所周知，具有相同辨别力和不同猜测参数的固定效应项目响应模型族没有模型可识别性。对于该族，将包括Fisher信息和新信息在内的某些类型的信息最大化以进行模型识别。给出了关于调谐参数的这些类型信息的单调性条件。在带有猜测参数的逻辑模型的情况下，表明在某些参数化下不存在最大值，其中可以使用负下渐近线而不改变考生正确回答的概率。