| Entrance psychological test has been applied in China for enlistees, in order to adapt to modern wars, improve army's performance, and prevent the young with psychological problems entering the service. Meanwhile, this program has posed press to the enlisting offices, for too many candidates of enlistees need to be tested. As a result, it has become a demanding task to shorten the testing time and improve the testing efficiency under proper validity. At the same time, with the development of psychological and educational measurement, item response theory (IRT) has been studied and applied in various areas to increase the testing efficiency and reduce measurement error. Additionally, computerized adaptive testing (CAT), which is based on IRT, has been used in many large-scale testing (TOFEL,GRE etc.) as a brief and valid method. So this study focused on applying IRT into Entrance Psychological Test for Enlistees program, evaluating its validity, and suggesting methods to optimize it. However, there is no successful IRT model for personality questionnaire, because of its complexity. Consequently, it is important both in theory and in practice to investigate feasibility and proper method of applying IRT into Chinese Soldier Personality Questionnaire (CSPQ). This study includes three parts. The first is applying 2-parametric Logistical Model (2PL) to analyze data of 2676 normal young people, 221 schizophrenia patients and 274 young people who have tendency of schizophrenia. Then compare the results with ones of classical testing theory. Second part is real-data simulation to discuss the probability of using Count-down method, which has been Successfully used in the CAT version of MMPI, to realize CAT of CSPQ. The third part is to develop CAT-CSPQ based on results of the second part, and put it into practice to validate. This study provides first step towards CAT of CSPQ, and puts forward the direction of further research.Here are the main results of this study:(1) In the first part, correlation between IRT scores and CTT scores didn't change obviously, before and after deleting the items with improper discriminating parameter a (out of the range of (0.3, 4)). This result indicated that the deleted items had contributed little to the estimation of subjects'abilities, and the model was robust.(2) In the first part, deleted items were mainly reverse-keyed items. The reasons may be that the reversed keying confused the subjects; that reversed keying made the locations of items extreme and lowered the discrimination parameters; that reversed keying made the item ambiguity; and that keying direction influenced the calculation of discrimination parameters. However, the real mechanism needs to be discussed further.(3) In the first part, distributions of the three clinical subtests under IRT and CTT changed obviously for schizophrenia group, while no significant changes were found for normal and disqualified groups. Distribution of schizophrenia group was closer with normal group under CTT scores, and closer with disqualified group under IRT scores. Additionally, distributions of three groups were more distinctive under IRT scores.(4) Results of part two and three showed that Count-down method can be used in the testing of CSPQ, and can improve the efficiency significantly.(5) Results of part two implied that testing efficiency can be further increased by ordering items from the least endorsed item to the most endorsed, and when the differences of endorsement was very small (below 0.01), put items with greater scalability coefficient H first, then the smaller ones.(6) In order to avoid influences of item-order rearrangement to the subjects'answers, items belonged to different subtests were rearranged as original order. Results showed that the efficiency was quite match to the best order.(7) Real testing showed that Count-down method can be used in the testing of CSPQ, and the scores didn't change obviously compared with scores derived from original scale. It can be used in the real Entrance Psychological Testing for Enlistees program.Conclusion: Applying IRT into personality questionnaire is rare in China, and is being discussed worldwide. This study applied 2PL to analyze the data, found out the advantages of IRT over CTT. The first is IRT can better discriminate the item contribution to the evaluation of subjects'abilities, and use fewer items to get more valid results. And the second is using IRT can differentiate subjects more clearly. However, due to the complexity of personality questionnaire, further research is still needed to completely use IRT into this area. And further researches are still needed to answer why most improper items were reversed keyed. As a simple approach to realize CAT of personality questionnaire, Count-down method can be used in CSPQ, and help achieve great efficiency. Additionally, Mokken Scale Model was introduced for the first time in this study, and used to determine the best item order of CAT-CSPQ. At last, new approach to realize CAT of CSPQ has been put forward as Pattern Analysis Procedure.
|
PDF Full Text Request