Imputing Prociency Data under Planned Missingness in Population Models

Abstract

From the perspective of missing data, in most assessments the proficiency of interest is a latent variable that is missing for all respondents. We never observe "intelligence" or "reading skill" directly; we only observe how respondents react to certain tasks that are assumed to be indicative of these underlying variables. The responses to these tasks, however, are fallible indicators at best, and only in naïve approaches to measurement are these directly equated to the underlying variable of interest.