Tuesday, November 9, 2021
2:00 – 4:00 PM (PDT) in BWW 1212 and Zoom:
Outline:
--- I will argue that psychometric models define the characteristics the data must possess to have a scale of measurement. In this sense, a model if prescriptive; it defines an ideal for measurement. This places goodness of fit research to the forefront of modeling; if the model does not fit the data, we do not have a scale.
--- Whether a model is useful in practical scale construction, depends on the cognitive or affective processes that lead a person to respond to items. Do these processes produce responses and data consistent with the model structure? To what degree can we "force" a model upon our data?
--- Ideally, a theory about the attribute (ability, trait, attitude) guides the operationalization of the attribute into a set of items that elicit the responses that are informative of the attribute. A model implied by the theory is a formal tool to test whether the predictions are correct. The outcomes provide feedback to the theory and can have implications for the items. In this conception, the model is subordinate to the theory and the data; it is a tool.
--- Is there enough sound theory about psychological attributes--educational measurement is a little different--to justify psychological measurement? Are there examples of successful theory and scale construction and what have they brought us? How should we continue?
Biography:
Klaas Sijtsma is a professor of methods and techniques of psychological research at Tilburg University, and specializes in psychometrics. His work focuses on measurement models and goodness of fit methods, but also on outliers and missing data, questionable research practices and the history of science. He published many articles and book chapters, and three monographs on test theory and psychometrics. He is a past President of the Psychometric Society.