Files
Abstract
The psychological response process is the judgement process that individuals undergo when responding to stimuli such as self-report items. The dominance assumption the most common assumption about response processes implies that individuals with higher standing on the trait continuum are more likely to endorse the item. On the other hand, the ideal point assumption implies that an individual is more likely to agree with the item when it is close to their standing on the trait continuum. Although the dominance assumption has been pervasive in psychological research, new evidence suggests the ideal point model may be more appropriate for self-report measures of individual differences, such as personality, attitudes, and interests. One major challenge faced by researchers is correctly understanding the response process to enable accurate theoretical and statistical inferences about psychological attributes. Inappropriately applying a dominance scoring to ideal point data can result in major substantive misinterpretations, including finding spurious dimensions in truly unidimensional data (Davison, 1977), and can lower power in tests for curvilinearity and interactions (Carter et al., 2017). Alternatively, inappropriate application of ideal point scoring to dominance data can result in high false positive rates for tests of curvilinearity and interactions (Carter et al., 2017). Currently, however, there is little guidance available on how best to determine whether responses arise from a dominance or ideal point response process, and the guidance that does exist is mixed. In two Monto Carlo simulation studies, the effectiveness of log-likelihood (LL), Akaike information criteria (AIC), Bayesian information criteria (BIC), M2 statistics, root mean square error of approximation (RMSEA), Tucker-Lewis index (TLI), comparative fit index (CFI), and adjusted 2/df ratios in identifying the correct item response process. Even though results showed that no one universal index can point to the correct response process 100% of the time, it is more appropriate to use the LL, AIC, and BIC to assist with such decisions.