Teacher effect change model: latent variable regression 5-level hierarchical model
- PDF / 773,243 Bytes
- 14 Pages / 595.276 x 790.866 pts Page_size
- 30 Downloads / 146 Views
Teacher effect change model: latent variable regression 5‑level hierarchical model Kilchan Choi1 Received: 13 February 2020 / Revised: 6 May 2020 / Accepted: 15 September 2020 © Education Research Institute, Seoul National University, Seoul, Korea 2020
Abstract This paper proposes a teacher effect change model in the form of a latent variable regression 5-level hierarchical model (LVR-HM5). Using multiple years of student achievement data, the LVR-HM5 attempts to simultaneously estimate teacher effect as well as teacher initial status and the gap parameter to model the change of such latent parameters over time. The gap parameter, the latent variable regression coefficient (Choi and Seltzer 2010; Choi and Kim 2019), captures the relationship between initial status and rates of changes within each year’s classroom. Furthermore, the LVR-HM5 allows us to model the teacher effect over time as a function of both time-varying and time-invariant characteristics. Such studies that focus on finding key correlates of teacher effect may have policy implications on teacher education, teacher professional development, and teachers’ instructional strategies that are potentially associated with improving teacher effectiveness.
Introduction During the last two decades, a number of states and school districts in the U.S. have revamped their teacher evaluation systems to incorporate components that are directly linked to student learning. Student learning is measured typically by an annual state assessment, and those achievement data are used for evaluating teacher and/or school effectiveness and accountability. While moving from research to practice, a number of unsolved methodological challenges have arisen. One such example is the lack of year-to-year stability of annual teacher value-added estimates. Research has shown considerable variation in teacher performance, as measured by value-added estimates, over time (Aaronson et al. 2007; Ballou 2005; Goldhaber and Hansen 2013; McCaffrey et al. 2009). Year-to-year correlations in value-added estimates in elementary and middle school mathematics teachers range from 0.2 to 0.5, and 0.3 to 0.7, respectively (McCaffrey et al. 2009). As these correlations appear to be low for a measure underlying high-stakes decisions, teacher evaluation systems that incorporate value-added estimates often require the inclusion of multiple years of growth data for students assigned to the teacher. The variation in the estimated
* Kilchan Choi [email protected] 1
CRESST/UCLA, Los Angeles, CA, USA
teacher effects on student performance include estimation error and other sources of nonpersistent variation in test performance, in addition to persistent differences in performance between teachers (Gordon et al. 2006). Under this circumstance, aggregating multiple years of teacher effects can help reduce potential bias in annual estimates caused by nonrandom teacher–student sorting and also improve the stability and predictive power of teacher effect measures (Ballou 2005; Goldhaber and Hansen 2013). As
Data Loading...
