Conférence-midi du GRIÉMÉtic
Par Carl Falk, professeur à l’Université McGill
(conférence en anglais)
Diffusion en ligne : https://zoom.us/j/692339850
Abstract:
The use of Likert-type items is pervasive throughout the social and behavioral sciences (e.g., rate your agreement to a statement on a scale from 0 – Disagree to 4 – Agree). However, individuals sometimes perceive and use the scale in different ways. In this talk, I present an item response theory model capable of modeling multiple response styles across multiple constructs of interest. Some features of this model include the ability for researchers to: 1) Flexibly determine how response styles are defined, 2) Test whether modeling particular response styles improves model fit, 3) Obtain scores for participants on the constructs of interest and response styles, and 4) Test for group differences in the response styles or substantive constructs. Comparisons with sum score-based approaches, and recent extensions to multilevel data will be briefly presented. Time permitting, additional challenges in aberrant (atypical) responding to survey items will be discussed.
Biography: Carl F. Falk is an assistant professor of Quantitative Psychology and Modeling in the Department of Psychology at McGill University. His work focuses on the development, testing, and computer programming of advanced latent variable models with applications across the social sciences (e.g., psychology, education, health outcomes, sociology, etc.). This work spans across item response theory, structural equation modeling, and multilevel modeling. Recent applied work includes analysis of field test data and simulation based evaluations of item banks for the Smarter Balanced Assessment Consortium.
