A Case Study of Some Practical Challenges of Longitudinal Student Achievement Modeling
The RAND Mosaic II Study
Published in: Longitudinal and Value Added Models of Student Performance / edited by Robert W. Lissitz (Maple Grove, Minn. : JAM Press, 2006), chapter 8, p. 230-254
Posted on RAND.org on December 31, 2005
The era of test-based accountability in education has spawned increased standardized testing and data collection efforts nationwide, raising both interest in opportunities for longitudinal modeling of student achievement. Longitudinal achievement trajectories of individual students represent a substantially richer information base than, for example, the longitudinal school-level aggregate measures of student achievement tracked by NCLB. Longitudinal analyses of individual student data thus provide numerous opportunities for educational research and evaluation. For example, there is considerable growing interest in so-called value-added modeling, where the achievement trajectories of individual students, coupled with histories of linkages to teachers and schools, might be used to identify effective teachers or schools (McCaffrey, Lockwood, Koretz, and Hamilton., 2003; Zvoch and Stevens, 2003; Lissitz 2005). Longitudinal student achievement information is also a growing contributor to the research base on effective teaching practices, teacher credentialing, and other educational interventions (Hanushek, Kain and Rivkin, 2002; Zimmer et al., 2003; Goldhaber and Anthony, 2004; Schacter and Thum, 2004; Bifulco and Ladd, 2005) where longitudinal information can lead to more precise and potentially less biased estimates of the effects of exposures to teacher or school-level predictors than would be possible with purely cross-sectional observational data. Finally, there is hope that individual longitudinal student achievement information might be used by educators in real time to assess individual needs and to adapt instructional practices accordingly (Supovitz and Klein, 2003; Gill et al., 2005).