Abstract
The concepts and methods of psychometrics originated under trait and behavioral psychology, with relatively simple data, used mainly for purposes of prediction and selection. Ideas emerged over that nevertheless hold value for the new psychological perspectives, contexts of use, and forms of data and analytic tools we are now seeing. In this chapter we review some fundamental models and ideas from psychometrics that are be profitably reconceived, extended, and augmented in in the new world of assessment. Methods we address include classical test theory, generalizability theory, item response theory, latent class models, cognitive diagnosis models, factor analysis, hierarchical models, and Bayesian networks. Key concepts are these: (1) The essential nature of psychometric models (observations, constructs, latent variables, and probability-based reasoning). (2) The interplay of design and discovery in assessment. (3) Understanding the measurement issues of validity, reliability, comparability, generalizability, and fairness as social values that pertain even as forms of data, analysis, context, and purpose evolve.
Original language | English |
---|---|
Title of host publication | Computational psychometrics |
Subtitle of host publication | New methodologies for a new generation of digital learning and assessment |
Editors | A.A. von Davier, R.J. Mislevy, J. Hao |
Publisher | Springer |
Pages | 81-107 |
ISBN (Print) | 978-3-030-74394-9 |
DOIs | |
Publication status | Published - 2021 |