Files
Abstract
Educators desire metrics to evaluate student progress over time. However, merely tracking changes in students' scores is often deemed insufficient. Stakeholders are also interested in evaluating whether individual student growth or class average growth is satisfactory or lacking compared to other students or classes. The student growth percentile (SGP; Betebenner, 2009) was devised to compare student growth to the growth of peers with similar score histories. SGPs are commonly used with assessments that provide sum or scaled scores generated from item response theory (IRT) models. However, such assessments fall short of meeting the increasing demand for reliable results that pinpoint students' specific strengths and weaknesses. This demand can be addressed through diagnostic assessments designed for use with diagnostic classification models (DCMs; Rupp et al., 2010). The recent transition of diagnostic models and methods from theoretical research to practical application makes it crucial that psychometricians provide metrics to evaluate growth within the DCM framework. Hence, this dissertation introduces the diagnostic growth percentile (DGP)—an adaptation of SGP for use with student results from DCMs—and reliability metrics for the DGP. Specifically, to evaluate the efficacy, reliability, and validity of DGPs, I conducted a simulation study and two empirical data analyses, which illustrate the computation, interpretation, and utility of the DGPs. Results of these studies showed that DGPs have acceptable levels of reliability for various assessment conditions and are viable approaches for comparing student growth in the DCM framework. I conclude this dissertation by comparing SGPs and DGPs. In sum, DGPs overcome some issues that plague SGPs while introducing new limitations that pose issues for potential uses and interpretations. This dissertation explores one approach for using an SGP-like metric in the DCM framework and shows favorable results for the DGP metric. However, additional research is needed before DGPs can be used in practice, and, as always, anyone interested in using DGPs in practice should prepare a comprehensive validity argument to support their intended use(s) for DGPs in their specific situations.