The concerns of this investigation are multiple sources of complications in the analysis of multilevel educational data. Selected multilevel methods provide some indication of misspecification and can identify the direction of the bias in estimating teacher/class effects on mean class outcomes. (Author/CTM)

The problem of choosing the most efficient design (longitudinal, mixed longitudinal, or cross-sectional) to estimate polynomial parameters for time-structured data in educational and psychological research is considered. The comparison of efficiencies shows that the assumed degree of the polynomial is crucial for selecting the most efficient…

To facilitate meta-analysis of diverse study findings, a mixed linear model with fixed random effects is presented and illustrated with data from teacher expectancy experiments. The standardized effect size is viewed as random and the variation among effect sizes is modeled as a function of study characteristics. (Author/BS).