全国地理奥林匹克竞赛与国际地理奥林匹克竞赛的试题类型、数量和特点相同，这些试题特征体现了地理学发展的趋势及对地理学科核心内容和关键能力的考查方式，有很多理念与我国地理课程改革要求契合。相对于国际地理奥林匹克竞赛，全国地理奥林匹克竞赛主要以中文形式进行考查，适合中国考生进行训练。若深入研究这些试题，可以有效地提高中学生的地理素质，推动高中地理教育的发展，促进高中地理课程的改革。国内试题的新要求与全国地理奥林匹克竞赛在学科核心素养、测试内容、具体任务、情境水平方面有很多相近之处。在全国地理奥林匹克竞赛中取得的经验，拓宽了地理学的视野，把握世界上先进的地理学思想，并为我国的高考命题提供了有益的对比与思考。

本研究通过Rasch模型对我国近十五年全国地理奥林匹克竞赛多媒体测试题和综合题进行分析，主要研究对象为12套多媒体测试题和12套综合题。首先，用Rasch模型对这些试题进行处理和分析，可获得信度、误差、分离度、单维性检验、试题-被试对应、数据模型拟合、点-测量相关等指标,归纳其主要特点，找到影响试题质量的主要问题；之后，尝试分析试题问题的形成原因，结合Rasch模型的各项数据量化该问题的影响因素，构建出解释模型。用SPSS软件进行单因素方差分析，对该模型进行修正；最终，综合前两部分研究内容，提出该问题的修正策略，筛选出数据表现好的试题构建题库，为后续培养学生奠定基础。本研究主要结论如下：

（1）利用Rasch模型，得到12套全国地理奥林匹克竞赛多媒体测试题和综合题的分析结果。经单维性和局部独立性检验，多媒体测试题和综合题都适合用Rasch模型进行分析。在具体分析中，两类试题的整体质量较好，绝大多数试题满足参数指标要求。但存在试题没有很好地将被试进行区分的现象。在多媒体测试题中存在较多难度过大和难度过小的试题，综合题中存在较多难度过大的试题。

（2）利用量化分析的研究方法，建立综合难度模型。以12套多媒体测试题和综合题中难度过大、难度过小和数据表现好的试题为例进行难度影响因素的分析。根据前人研究，结合地理情境水平、地理学科能力水平等构建出符合地理奥赛特点的综合难度模型。全国地理奥林匹克竞赛多媒体测试题和综合题的综合难度模型由推理水平、学科能力水平、情境水平三维度组成。图像数量和知识含量与试题难度无明显对应关系。在多媒体测试题中，就三个影响因素与难度的相关程度来说，综合难度的第一影响因素为推理水平，第二影响因素为情境水平，第三影响因素为学科能力水平。在综合题中，综合难度的第一影响因素为情境水平，第二影响因素为学科能力水平，第三影响因素为推理水平。本研究计算出数据表现好、难度过大、难度过小的试题综合难度值，在之后命题时可参考这三个数值对试题的前期难度进行了预测，以便更准确、科学地测试学生。

（3）基于前文的研究成果，归纳试题难度的调控策略和题库构建策略。建议在修改难度过大和难度过小试题时，根据具体题目进行调控。本文提出调整试题推理水平、增减试题的学科能力水平、改变试题的情境水平这三个调控措施，具体实施时要考虑题意和考生作答情况进行修改。对于数据表现好的试题，建议构建题库。本次研究中可纳入初级题库的试题有多媒体测试题322道，综合题227道。若后期能将其他指标异常题目进行修改，经等值化处理后也可纳入题库。在具体应用题库时，可从12个地理奥赛内容主题角度、考生能力水平角度进行试题筛选和重组，从而较为准确和全面地培训考生。

The type, quantity and characteristics of the test questions of the National Geography Olympiad and the International Geography Olympiad are the same. These test questions reflect the development trend of geography and the examination methods of the core contents and key abilities of geography. There are many ideas in line with the requirements of China's geography curriculum reform. Compared with the International Geography Olympiad, the National Geography Olympiad is mainly tested in Chinese, which is suitable for the training of Chinese candidates. If we deeply study these questions, it can effectively improve the geographical quality of middle school students, promote the development of high school geography education and promote the reform of high school geography curriculum. There are many similarities between the new requirements of domestic test questions and the National Geographic Olympiad in terms of discipline core literacy, test content, specific tasks and situational level. The experience gained in the National Geography Olympiad has broadened the vision of geography, grasped the advanced geographical ideas in the world, and provided useful comparison and thinking for the proposition of college entrance examination in China.

    This study analyzes the multi media tests and written response tests of National Geography Olympiad in recent 15 years through Rasch model. The main research objects are 12 sets of multi media tests and 12 sets of written response tests. Firstly, these questions are processed and analyzed with Rasch model, and the indexes such as reliability, error, separation, one-dimensional test, question subject correspondence, data model fitting, point measurement correlation and so on can be obtained. Summarize its main characteristics and find the main problems affecting the quality of test questions. After that, try to analyze the causes of the test questions, combined with the data of Rasch model, quantify the influencing factors of the problem, and build an interpretation model. SPSS was used for one-way ANOVA to modify the model. Finally, integrating the first two parts of the research content, this paper puts forward the correction strategy of the problem, selects the test questions with good data performance, and constructs the question bank, so as to lay the foundation for the subsequent training of students. The main conclusions of this study are as follows:

(1) Using Rasch model, the analysis results of multi media tests and written response tests of 12 National Geography Olympiad are obtained. Both single dimension and partial dimension were tested by Rasch model. In the specific analysis, the overall quality of the two types of test questions is good, and most of the test questions meet the requirements of parameter indicators. However, there is a phenomenon that the subjects are not well distinguished in the test questions. There are many questions with too much difficulty and too simpleness in multi media test questions, and there are many questions with too much difficulty in written response tests.

(2) Using the research method of quantitative analysis, a comprehensive difficulty model is established. Taking 12 sets of multi media tests and written response tests with too much difficulty, too little difficulty and good data performance as examples, the influencing factors of difficulty are analyzed. According to previous studies, combined with the level of geographical situation and the level of geographical subject ability, this paper constructs a comprehensive difficulty model in line with the characteristics of Geography Olympiad. The comprehensive difficulty model of multi media tests and written response tests of National Geography Olympiad consists of three dimensions: reasoning level, subject ability level and situation level. The number of images and knowledge content have no obvious corresponding relationship with the difficulty of the test questions. In multi media test questions, in terms of the correlation between the three influencing factors and difficulty, the first influencing factor of comprehensive difficulty is reasoning level, the second influencing factor is situational level, and the third influencing factor is subject ability level. In the written response tests, the first factor affecting the comprehensive difficulty is the situation level, the second factor is the subject ability level, and the third factor is the reasoning level. This study calculates the comprehensive difficulty value of the test questions with good data performance, too great difficulty and too small difficulty. In the later proposition, we can refer to these three values to predict the early difficulty of the test questions, so as to test the students more accurately and scientifically.

(3) Based on the previous research results, this paper summarizes the control strategy of test question difficulty and the construction strategy of question bank. It is suggested to adjust and control according to the specific questions when modifying the questions that are too difficult and too difficult. This paper puts forward three control measures: adjusting the reasoning level of the test questions, increasing or decreasing the subject ability level of the test questions, and changing the situation level of the test questions. The meaning of the questions and the answers of the candidates should be considered in the specific implementation. For the test questions with good data performance, it is suggested to build a question bank. In this study, 322 multi media test questions and 227 written response tests questions can be included in the primary question bank. If other index abnormal questions can be modified later, they can also be included in the question bank after equivalence treatment. In the specific application of the question bank, the test questions can be screened and reorganized from the perspective of 12 Geographical Olympiad content themes and candidates' ability level, so as to train candidates more accurately and comprehensively.