Investigating the Effect of Differential Rapid Guessing on Population Invariance in Equating

Score equating is an essential tool in improving the fairness of test score interpretations when employing multiple test forms. To ensure that the equating functions used to connect scores from one form to another are valid, they must be invariant across different populations of examinees. Given that equating is used in many low-stakes testing programs, examinees' test-taking effort should be considered carefully when evaluating population invariance in equating, particularly as the occurrence of rapid guessing (RG) has been found to differ across subgroups. To this end, the current study investigated whether differential RG rates between subgroups can lead to incorrect inferences concerning population invariance in test equating. A simulation was built to generate data for two examinee subgroups (one more motivated than the other) administered two alternative forms of multiple-choice items. The rate of RG and ability characteristics of rapid guessers were manipulated. Results showed that as RG responses increased, false positive and false negative inferences of equating invariance were respectively observed at the lower and upper ends of the observed score scale. This result was supported by an empirical analysis of an international assessment. These findings suggest that RG should be investigated and documented prior to test equating, especially in low-stakes assessment contexts. A failure to do so may lead to incorrect inferences concerning fairness in equating.