A Study On The Factors Of The Difficulty Of Multiple-choice Questions Of Senior School

The difficulty is one of the very important aspects to evaluate the test quality. In the general study, the difficulty of a question is the ratio of the number of the people who are right to all. But, for originators, they can’t decide whether to adopt the title by a test first. So far, to estimate difficulty through the experience method is usually uncertainty and subjectivity. Therefore if we can obtain a set of scientific prediction difficulty formula or models, which can accurately predict the difficulty of questions before a test, it will improve the quality of test papers.The main factors affecting test-question’s difficulty include the natural factors, test level and other factors. In the study, we strictly control the test level and other interference factors by select data sample. Focus on analyzing the natural factors of the difficulty of multiple-choice questions. Through the literature analysis and investigation,we bring forward the hypothesis of ten affecting factors affecting difficulty of test-questions, that is " situation or context","number of knowledge module","number of knowledge module","number of knowing-level checkpoints","number of comprehending-level checkpoints","number of analyzing-level checkpoints","number of checkpoints of low-level operation","number of checkpoints of primary level operation","number of checkpoints of advanced level operation","number of solving methods","prompt level of options".Draw lessons from domestic and foreign statistic case, the study carry on the quantification analysis to each factor. Statistically, there are four factors which significantly affect the difficulty of test-questions i.e.," situation and context","number of knowledge module","number of comprehending-level checkpoints","number of analyzing-level checkpoints", and correlation coefficients are0.520,0.494,0.732,0.788in turn. Then, we analyze all the result qualitatively to attempt to find out more important information for teaching and setting a paper.Finally, we take a simple comparison of two approach of predicting the difficulty of test-questions, which are Artificial Neural Network (ANN) and Multiple Regression Analysis. Based on Multiple Regression Analysis, we gain a estimating-model for difficulty:y=0.274x1+0.081x2+0.142X3+0.087x4-0.198. Its correlation coefficient is0.860, R Square is0.739.Although the result is not perfect, the study provides a method of predicting the difficulty of multiple-choice question system, with some reference value.