Publication details
Modeling Extreme Response Style Using Item Response Trees
| Authors | |
|---|---|
| Year of publication | 2017 |
| Type | Appeared in Conference without Proceedings |
| MU Faculty or unit | |
| Citation | |
| Description | It is well known that responses to Likert scale items are potentially influenced by response styles and not only by the latent trait level or attitude strength to be measured. Here, we focus on the extreme response style (ERS) in responses to a Big Five personality test. The purpose of the two presented studies (N = 7885 and 7623) is to demonstrate an attempt to jointly model the latent personality traits and ERS using a class of Item Response Theory models - the Item Response Trees - such that the extreme response style effect can be measured and removed. The Item Response Trees framework can be utilized to decompose item responses into a priori defined sub-responses or hypothesized response processes and model them as functions of different latent variables. The results indicate extreme response style can indeed be modeled and separated from the substantive trait(s) in this way, and represent important findings for the practice of using response scales. Further applications to measurement of ERS and other response styles are discussed. |