Using Item Response Theory (IRT) Methods to Refine Measurement & Reduce Participant Burden in Aging Research
January 26, 2024
Abstract: Item Response Theory (IRT) methods are best known for their role in large-scale achievement tests, such as the SAT and GRE. However, IRT methods have unique—and often overlooked—advantages for improving measurement and reducing participant burden in aging, and related fields. The first advantage of IRT is illuminating how precise a measure is for people across different levels of the underlying latent construct, an affordance not provided by factor analysis. This more precise capacity to estimate participants’ level of ability better-informs identification of selective attrition in panel surveys, such as whether low-ability respondents were more likely to attrit. The second advantage of IRT is identifying redundant and/or imprecise items for removal, reducing longer scales in favor of more efficient, yet similarly precise, streamlined measures. IRT is therefore useful in reducing the time and money it costs to administer surveys.
These two advantages are also important in reducing participant burden. Shorter measures may increase consent rates by reducing participant time burden, which may lead respondents to consent to subsequent data collections. Shorter measures may similarly minimize attrition, in that participants are more likely to complete shorter surveys – which also have the effect of yielding better data quality. In sum, by producing shorter and more efficient measures, IRT has the potential to reduce attrition, increase consent rates, and improve data quality in aging research.
Additional Information: This introductory workshop minimizes notation and instead emphasizes conceptual understanding and application, with a particular emphasis on figural representations of IRT. The Graded Response Model, which is designed for analyses of Likert-type items (e.g., strongly disagree to strongly agree), will be the primary focus.
The instructor reviewed syntax and interpreted output and figures in R (mirt package) and in MPlus. This free workshop was presented as part of the hybrid 2024 NIMLAS Annual Plenary Meeting; participants joined in-person at ISR and online via Zoom.