The Primary Development Of Arithmetic Reasoning Item Based On IRT For Citizen Recruited

With the universality of calculator technique and the development of item response theory (IRT), the computerized adaptive testing (CAT) just and increasingly becomes the mainstream of international large tests. Because the CAT has many advantages, such as to save time, much safety and to provide the results of examinees in time, etc., it is applied to the army's selection and classification quickly. Actually, all CAT tests are based on IRT. Currently, citizen recruited psychological screening system is an election system for the recruitment in order to prevent the retarded and the schizoid entering into troop. However, with the demanding of army modernization, the military personnel's classification and the match of personnel and allocation have already become the necessary step that may speed the course of quality model restructuring, which puts forward a higher request to screening tests. The CAT based on IRT is the most ideal large-scale test form. Under this background, this investigation just puts forward an assumption to discuss the measuring instrument for arithmetic reasoning ability of citizen recruited. To develop items with IRT maybe lay on the foundation for the establishment of citizen recruited storage, and especially provide the possibility to develop the computerized adaptive testing ability.Main content and procedure are as follows:1. Determine the research thought and put forward the theory assumption through literature review and consultation of experts;2. Develop the arithmetic reasoning items with IRT;3. Test 1047 recruits with the primary arithmetic reasoning subtests and carry on the item analysis;4. Edit the subtests and test 3655 recruits to collect the actual data;5. The item analysis method includes: The principal components analysis was carried out to measure the unidimensionality of the data. Three-parameter logistic model was used to fit the data collected. The item parameters were evaluated by Bayesian expected posteriori estimation with BILOGMG software.χ2 adjusted assessment was used to assess model-data fit. The research adopted anchor-item design and finished test equating with the item characteristic curve equating to select the more effective items. Test information function was used to determine standard error of estimation;6.Verify the theory assumption.Main results of this study are as follows:1. Theory assumption, the adult's (citizen recruited) mathematics cognition level experienced from the addition structure to integral multiplication structure, to fractional multiplication structure, to the initial construction of direct proportion concept, to the initial integration of direct and inverse proportion concept and to high level integration of direct and inverse proportion concept, which is formed by some ordinal phases, such as from low-class phase to high-class one, from simple phase to complicate one, from unidimensional phase to multidimensional one, etc., is proved practicable;2. According to the theory assumption of the adult's (citizen recruited) mathematics cognition level, we wrote item manual and retained 134 arithmetic reasoning items, which capabilities were good. 4 subtests were in accordance with the request of unidimensionality. The data fitted three-parameter logistic model, AdjChi/df≤3, 0.5