Complex Applications of HLM in Studies of Science and Mathematics Achievement: Cross‐Classified Random Effects Models

Hierarchical linear models have become a familiar method for accounting for a hierarchical data structure in studies of science and mathematics achievement. This paper illustrates the use of cross-classified random effects models (CCREMs), which are likely less familiar. The defining characteristic of CCREMs is a hierarchical data structure defined by multiple random factors at higher levels. We illustrate CCREMs using data for approximately 10,000 students from more than 250 high schools who attended one of 27 four-year postsecondary institutions. The appropriate use of CCREMs helps to ensure unbiased estimates of effects and credible inferences.