Abstract

This chapter discusses the general concepts of identification and partial identification of statistical models. We elucidate the identification restrictions to endow with meaning the parameters of interest of the fixed-effects one-parameter logistic model with guessing (1PL-G), a model used in educational measurement. We also review the restrictions for identifying the average treatment effect (ATE) in evaluating a policy or program. To address the fundamental problem of causal inference, we also present a partial identification analysis of the ATE. On the basis of the results, we emphasize the relevance of an identification analysis and the usefulness of considering a partial identification approach in causal inference.