Multiple Comparisons And Their Applications To Evaluation Of Education Quality Of Universities

Multiple comparisons are important test methods. In the past two decades, researches and applications of multiple comparisons have rapidly developed, particularly in the areas of industries, biology and clinical trials, and many other areas. This paper is broken into three parts:introduction to the construction of the joint confidence intervals, the error rate control, and the application of multiple comparisons to the teaching evaluation in the case of Chizhou College.Firstly, we investigated the structure on the joint confidence interval of multiple comparisons by comparing the structure and the excellence of the joint confidence interval. In the constructing of the joint confidence interval, the current investigations are mainly aimed at the general case of the binary normal situations, the Bonferroni joint confidence intervals, and the Scheffe joint confidence intervals. In this thesis, we not only simplified the calculation quantity and made it extend to the case of multivariate normal situations, but also compared the excellence of the new joint confidence intervals with the classic that.Secondly, we optimized the mistakes rate control of multiple comparisons. In this aspect, most scholars are involved in the study on the classical FWER. However, the condition of the classical FWER is very severe. Therefore, this thesis focuses on FDR. We firstly reviewed the control process of the classical BH, and then gave several higher optimization processes in effects, and finally validated them by Matlab.Thirdly, the method of multi-comparisons was applied to the teaching evaluation of Chizhou College. So far, the teaching evaluation of Chizhou College not only is simplex and irrational, but also does not have the correlative theoretical achievements. On the basis of the actual situation of mathematics in Chizhou College and a combination of the principal component analysis in mathematical statistics and the impact factor analysis, we established a suitable evaluation function for Chizhou College in this thesis. Using the method of multi-comparisons, we comparatively investigated the new evaluation results by the level of age, and clarify the influencing factor of different levels with evaluation results.In fact, there are some limitations in the application of multiple comparisons. If the quantity of sample is too large, or the frequency discrepancy repeated test under many levels is distinctly different, it will lead to enormous calculation quantity as well as the increase of uncertainty on the control of the first type of error.