摘要
早期的单维IRT模型忽视了测验多维性的可能,而多维IRT模型对各维度的划分不够明确,不能很好地反映各维度能力的内涵。高阶IRT模型承认测验的多维性,以分测验划分维度,同时又将多个维度的能力统一到一个高阶的能力中,能够在了解被试各维度的能力同时,为被试提供整体的能力估计,它能更好地反映实际,并且适应大规模测验的需求。
Item Response Theory （IRT） is a well known theory which can reflect the relationship between latent trait and items. However, both uni-dimensional IRT model and multi-dimensional IRT model are short at reflecting ability structure precisely and appropriately. The Higher-Order IRT （ H-O IRT） model is a muhi-unidimensional model that uses in-test collateral information and represents it in the co-relational structure of the domains through a higher-order latent trait formulation. H - O IRT model can provide both domain and the whole information, which is better fulfilling the needs of large-scale tests.
出处
《考试研究》
2014年第2期59-63,共5页
Examinations Research
