Abstract:
|
In low-stakes assessments, test performance has little or no consequence for examinees, who may hence not fully engage when answering items: Instead, disengaged examinees may guess or omit responses. Examinee disengagement poses a severe threat to the test validity of low-stakes assessments if not modeled properly. Statistical modeling approaches have been proposed separately for nonresponse or for guessing, but do not consider both types of disengaged behavior simultaneously. We combine research on modeling examinee engagement and research on missing values and present a hierarchical latent response model for identifying and modeling the processes associated with examinee disengagement jointly with the processes associated with engaged responses. We use a mixture model identifying disengagement by assuming different data-generating processes underlying item responses and omissions, respectively, as well as response times associated with engaged and disengaged behaviors.
|