Old Enough to Perceive Things Differently? Detecting Measurement Invariance Across Age Groups Using Item-Focused Tree

Abstract

Abstract The current study applies a new technique, item-focused tree (IFT), to investigate measurement invariance (MI) across age groups. MI is violated when differential item functioning (DIF) is detected. Compared to traditional DIF techniques, IFT does not require researchers to artificially split respondents into multiple groups when the covariate is continuous (e.g., age). Moreover, with IFT, researchers can examine DIF induced by multiple covariates at one time. The current study applies IFT technique to better understand how items in various measures may be perceived differently by respondents of varying ages, where IFT examined DIF induced by sex, work experience, and age simultaneously. Results from a two-wave dataset suggested that age was the sole covariate associated with all the DIF identified. Specifically, across both waves, older workers were more likely to express disapproval toward a few negative stated items (e.g., not learning, not picking up new skills, and having no positive effect on others), with moderate effect sizes. The DIF tree plots built by IFT algorithm further indicated the age of 50 years old as the key point to split 2 age subgroups where DIF is mostly like to be observed. Additionally, education was found to induce DIF jointly with age in the exploratory analyses. Implications, limitations, and future directions are discussed. A brief tutorial on how to conduct IFT using R is also included.