The Relationship Between Reaction Time And The Level Of Personal Traits And Person-item Distance

The computerization of personality test allows researchers to obtain response time information easily, and researchers do want to make a combination of item responses and response times to estimate the level of respondents’ trait with improving the accuracy of estimation. In order to lay the foundation for developing response time model of the personality test, this study examined the relationship between level of personal traits and its function and response time by using Eysenck Personality Questionnaire(EPQ) which based on unfolding models and cumulative models.This research adopted E-PRIME software to develop the EPQ test program. 237 respondents aged 18 to 26 volunteered to join the test. In the test, they were asked to keep the speed of their habits, and respond to the items according to the instructions shown on the computer. After the data collection, the researcher first used factor analysis to inspect the dimensions of two-scale of the test, and the two-scale was the extraversion(E) and neuroticism(N) scale. Then for the measurement data of these two scales, margin maximum likelihood estimation(MMLE) and markov chain monte carlo(MCMC) method were used to estimate the two-parameter logistic model(2PLM) and generalized graded unfolding model(GGUM) respectively. The two model-data fit were examined and compared by adjusted χ2/df ratios of item singles, item pairs,item triples. Finally, the product-moment correlations were calculated between the log-response times and the person-item distances within each item and within the whole test respectively.The results of the factor analysis showed that the first and second eigenvalue ratio of the E-scale was 2.662, and the N-scale was 3.763, which meant that N-scale met the unidimentional standard better than E-scale. On the E-Scale and N-scale, the adjusted χ2/df ratios of item singles,item pairs, item triples of 2PLM and GGUM were less than 3. On the E-scale, the model-data fit of generalized graded unfolding model were better than that of two-parameter logistic model, but on the N-scale, the opposite was true.The correlation between the level of personal traits and response time(or logarithmic response time) with two models of 2PLM and GGUM and two estimation methods of MMLE and MCMC got similar results. Namely, the relationship between the level of personal traits and response time, which was characterised by an inverted-U effect,only appeared on some items and scale, and the effect sizes was very small. Also, the correlation between the person-item distance and response time(or logarithmic response time)(DD hypothesis) also showed similar patterns, not met all items or scales.Therefore, this study find that we need to collect more data to test the role of response time in the parameter estimation of personality test.