Publication details
Reliability models in classical test and latent trait theories
| Authors | |
|---|---|
| Year of publication | 2021 |
| MU Faculty or unit | |
| Citation | |
| Attached files | |
| Description | Almost 120 years after Spearman's (1904) paper establishing foundations for the theory of measurement (which is nowadays known as the Classical Test Theory), a concept of "reliability" seems to be well understood. Lee J. Cronbach had a similar hunch in the 60th, which resulted in the generalizability theory, on which he was working the next ten years. More advanced estimators of reliability, based on factor analysis, such as omega coefficients, emerged during the following years. I believe the concept of reliability is misinterpreted and confused by a gross amount of different coefficients. Therefore, I will describe the epistemological foundation of the reliability theory (internal consistency). I will explain why the reliability defined as parallel tests (or "dimension-free" reliability) is not the same concept as reliability defined as explained variance ("model-based" reliability), or why and how the reliability in norm-referenced and criterion-referenced tests differs. Finally, I will provide several guidelines for estimator choosing based on the sample size, nature of measured attribute, dimensionality, or testing purposes. |