grmRT: An R Package for Joint Modeling of Likert Responses and Response Times Under the Distance-Difficulty Hypothesis

Computer-administered questionnaires record item response times, yet these timings rarely inform scale evaluation. We introduce grmRT, an R package that jointly models Likert responses with the graded response model and log response times with an inverted-U function of the person-item distance (the distance-difficulty hypothesis). The package estimates the model by marginal maximum likelihood, offers Student-t errors for noisy web-survey timings, and tests the distance effect item by item as a diagnostic of item functioning. In 11,605 respondents to the IPIP Big-Five markers, 48 of 49 timed items showed the expected pattern; the exception, “I insult people,” was among the least discriminating items.