This National Institute of Education report discusses problems of data aggregation and disaggregation which arise whenever the unit on which variables are measured is different from the unit of analytical, conceptual or policy-related interest. For example, researchers may want to know the effects of family background, of peer groups, or of school desegregation on individual students' achievement. Relevant data, however, may only be available for schools, even though the unit of analysis is the student. In situations such as this, the researcher must use findings based on the behavior of aggregates to make inferences about the behavior of individuals. This paper presents three methods for attempting to retain the advantages of aggregate data while minimizing the disadvantages. Detailed explanations of each of the three methods are provided along with a reanalysis of data from the Coleman study using some of the techniques presented. (Author/JKS)