In value-added modeling, where estimates of teacher fixed effects may be used for high stakes decisions, it is critical that analysts understand precisely how results can be applied and what parameters are being estimated by different routines. Large multi-level longitudinal student achievement databases with students linked to teachers are now available for many school districts and states. The availability of these data coupled with a desire to measure teacher performance has sparked great interest in estimating panel data models with one or multiple levels of fixed effects to produce value-added measures of teacher effectiveness. There are now a variety of standard and user-written Stata routines available to estimate such models, and center affiliates Daniel McCaffrey and J.R. Lockwood of RAND Corporation, along with Kata Mihaly and Tim Sass, have authored a forthcoming paper that promises to clarify interpretation of them.

Value-Added Modeling with Fixed Effects: A Review of Stata Routines for Linear Models uses straightforward examples to add clarity to estimated effects reported by various Stata fixed-effects modeling routines that are commonly used in value-added modeling. The different routines also use different computational approaches that can significantly affect the time and resources required to estimate effects from large datasets like those from states or very large school districts.