An Application of Hierarchical Linear Modeling to the Estimation of School and Teacher Effect	William J. Webster, Robert L. Mendro Timothy H. Orsak, and Dash Weerasinghe Dallas Independent School District

Figure 1 shows the nature of the equations used in the generation of school effectiveness indices. Each outcome variable is described under "Outcome" along with the grades at which it is included, the score that is the basis for the analysis, the methodology utilized, the level at which the data are analyzed (student or school level), possible predictors and the grades at which they are found, and the school level conditioning variables included in the student level equations. Two different regression models are used depending on whether the unit of analysis is the student, in which case hierarchical linear modeling is used, or the school, in which case multiple regression analysis is used. Through these approaches it is possible to obtain extremely reliable predictions of student and school outcomes and to compare actual outcomes to those that are predicted. All analyses that are done at the student level are calculated on residuals, that is, statistics that have had individual student characteristics over which the schools have no control removed from the equations (gender, ethnicity, limited English proficient status, socioeconomic status, and all of the interactions between those variables).

Classroom effectiveness indices are computed utilizing student and classroom data (classroom level conditioning variables) for the Iowa Tests of Basic Skills, the Tests of Achievement and Proficiency, the Texas Assessment of Academic Skills, the Texas Assessment of Academic Skills-Spanish, the Spanish Assessment of Basic Education, the Assessments of Course Performance, and the Woodcock-Muñoz Language Survey.