| Guanglei Hong
University of Michigan
Causal inference for multi-level observational data with applications to educational research
The purposes of this dissertation are to extend the conceptual framework of causal inference to encompass multi-level longitudinal educational data and to apply the extended framework to two important educational policy issues: kindergarten retention and within-class ability grouping. The dissertation starts with an investigation of the implications of this theoretical extension for various propensity-based causal inference techniques. This is to be followed by applications of the causal inference concepts and techniques to the empirical studies with the ECLS-K data. For the kindergarten retention study, the major causal question is how much more or less the kindergarten retainees would have accomplished in literacy and mathematics learning had these students been promoted to the first grade instead. In the second application study, I investigate the short-term and long-term causal effects of within-class homogeneous ability grouping in elementary reading instruction on children's literacy growth. In particular, I test the hypothesis regarding the widening of achievement gap between students in higher and lower reading groups as a result of within-class ability grouping. These empirical questions are not only of substantive interest to educational policy making but also representative of the multi-level designs that pose methodological challenges to causal inquiry.
Back to Funded Dissertation Grants Page |