Theory And Application Of Quantitative Reasoning Items Automatic Generator With Diagnosis Function

The purpose of the research is to develop an automatic quantitative reasoning item-generator using item generation method. This automatic generator is capable of generating item models with different difficulty values, based on cognitive design system method and using item-model method. The generator not only can generate items in a function level, but also has the capability of primary cognitive diagnosis by analyzing cognition processes and facts, and by establishing cognitive features'models.This paper reviews research processes, general methods and development phases of item generation. The theoretic and practical advantages of this advanced psychological testing method are also described. Quantitative reasoning is a kind of reasoning process based on quantity relation, which is usually evaluated by Algebra reasoning test. The solution process can be divided into five phases: problem translation, problem understanding, mathematic manipulation, process monitoring and solution execution. Problem-solving strategies and item stimulating features are also described. Traditional intelligent test can only evaluate subject's ability roughly, instead of their internal cognitive structure. Cognitive diagnosis, which is based on both cognition psychology and psychological testing, will help to find out subject's cognitive characteristic. Different diagnosis test needs to be analyzed by different psychological testing model, LLTM is used to establish quantitative reasoning item's cognitive feature model.Variable type, driving direction, problem type and number of hidden constraints are considered as four stimulating factors of quantitative reasoning items, and each of these factors are further divided into several levels. 20 items are produced with experimental design method. These items are tested on 6-grade-students. Item stimulating factors and item difficulty values are used to establish the regression equation. The equation is effective, and its determination coefficient is .848. Same direction, unknown variable type and constraint quantities are effective at the level of .01.Theoretically item structure can be divided into three parts: fixed content, content that affects difficulty and content that does not affect difficulty. Item generator is programmed by Visual Basic. Items can be produced by specifying either item conditions, or item difficulty. Difficulty prediction level and reliability of items produced by generator are .850 and .916 respectively. Item generator has primary cognitive diagnosis function.