Abstract: Hierarchical linear model (HLM) is an advanced statistical method for analyzing hierarchical structure data, and item response theory (IRT) is a modern measurement theory for accurately estimating individual’s ability. Combining HLM and IRT brought about the development of multilevel item response theory (multilevel IRT), where ability parameter in item response model is treated as latent variable in HLM. Multilevel IRT can be used not only to obtain the values of item parameters and ability parameters of different levels, but also to estimate the regression coefficients and variance components more precisely. In this paper, the authors expound the developmental history of multilevel IRT, then evaluate the academic value of multilevel IRT from the perspectives of differential item functioning, test equating and school effectiveness research. Future research directions for multilevel IRT are discussed as well.

Key words: multilevel IRT, item response theory, hierarchical linear model