Multifacet Universes of Admissible Observations and G Study Designs

Abstract

Often, generalizability analyses may be viewed as two-stage processes. The goal of the first stage is to obtain estimated variance components for a G study design, given a universe of admissible observations. The second stage involves using these estimated variance components in the context of a D study design and universe of generalization to estimate quantities such as universe score variance, error variances, and coefficients. This dichotomy of stages is somewhat arbitrary at times, but it is a conceptually meaningful distinction. Therefore, this chapter treats multifacet G study designs and universes of admissible observations, while Chapters 4 and 5 treat multi-facet D study design considerations and universes of generalization.