ABSTRACT
Most educational research data have a multilevel structure. For example, the data collected may be nested in students as a second level, students in classes as a third level, classes in schools as a fourth level, and so forth. Such structures require multiple levels of analysis to account for differences between observations, students, and other higher-level units. Since the typical clustering of multilevel data leads to (marginally) dependent observations, separate linear analyses based on the assumption of independent identically distributed variables at each of these levels are inappropriate.
