A Construct Modeling Approach to Measuring Fidelity in Data Modeling Classes

Abstract

In program evaluation research, measures of realized classroom instruction are often referred to as fidelity measures. Although there is a wide consensus that fidelity measures should be grounded in the program theories guiding the intervention, there is very little explicit discussion of how to adequately represent program theories, or how to scale a measure that can be interpreted in terms of the program theories. This dissertation is an example of a construct modeling approach to fidelity measurement. Here program theory is represented as the structure, processes, and underlying constructs of the designed intervention. Observable variables were generated and scored, and the data was modeled using a Partial Credit Model. The model largely supports the distinctions in the construct map and the correspondence between construct and scale. Additional implications for observation measures of classroom interactions are discussed.