S. Paul Wright

Teacher and Classroom Context Effects on Student Achievement: Implications for Teacher Evaluation

The Tennessee Value-Added Assessment System (TVAAS) has been designed to use statistical mixed-model methodologies to conduct multivariate, longitudinal analyses of student achievement to make estimates of school, class size, teacher, and other effects. This study examined the relative magnitude of teacher effects on student achievement while simultaneously considering the influences of intraclassroom heterogeneity, student achievement level, and class size on academic growth. The results show that teacher effects are dominant factors affecting student academic gain and that the classroom context variables of heterogeneity among students and class sizes have relatively little influence on academic gain. Thus, a major conclusion is that teachers make a difference. Implications of the findings for teacher evaluation and future research are discussed.

Two- and Three-Year Achievement Results From the Memphis Restructuring Initiative

The purpose of this study was to analyze student achievement data from the first 3 years of the Memphis Restructuring Initiative (MRI). The MRI represents one of the first efforts by an urban school district to move past traditional top-down versus bottom-up reform debates by providing systemic support for outside-in/inside-out implementation and local co-construction of externally-developed reform designs in schools. Analyses of academic achievement focus on a state-of-the-art measure of value added assessments. At the end of 3 years the reforming schools had produced generally positive gains relative to locally matched control schools. Those results varied somewhat by reform type and by level of poverty in the communities being served. Based on the research methods used and the results, implications for future research and practice in educational reform are discussed.

Inside Systemic Elementary School Reform: Teacher Effects and Teacher Mobility

In the 1995–96 and 1996–97 school years, 37 elementary schools in Memphis, TN began implementation of 1 of 8 comprehensive school reform designs. The effectiveness and mobility of teachers at these schools were examined longitudinally relative to teachers at 63 nonrestructuring schools. Analyses of teacher effectiveness scores, derived from student “value-added” achievement scores, indicate that the 1995 reform cohort only showed significantly greater gains in effectiveness relative to the nonrestructuring group. This pattern was most strongly pronounced for highly experienced teachers. Teacher mobility was found to increase minimally as a function of restructuring, largely due to district policy changes.

Repeated Measures Analysis Using Mixed Models: Some Simulation Results

Currently available software (e.g., the MIXED procedure in SAS) makes it easy to analyze repeated measures data having such “nonstandard” features as incomplete observations and structured covariance matrices. Unfortunately, in small samples, the hypothesis tests based on the asymptotically valid Wald chi-square statistics and related F statistics can be grossly inaccurate. Previously reported simulations showed that accurate results could be obtained from adjusted F statistics using both univariate (Greenhouse-Geisser and Huynh-Feldt) and multivariate (Lawley-Hotelling trace) approaches to adjustment. Both approaches worked well when using the potentially time-consuming REML estimates of covariances, but not with the faster, non-iterative MIVQUEO estimates. The present paper reports encouraging results from new simulations comparing REML with alternative non-iterative SSCP estimates based on sums-of-squares-and-cross-products of residuals — in effect, using 2-stage least squares.