| Educational measurement and teaching evaluation are the important parts of teaching activities. The main content of teaching evaluation is to analyze examination results and their influential factors. Usually, the statistic methods are as follows: descriptive analysis, cluster analysis, principal component analysis, factor analysis and multiple linear regression analysis. However, there are data with hierarchical structure in teaching evaluation, such as, students' scores in different schools or with different majors in the same school. The above-mentioned methods are not appropriate for analyzing such data, while multilevel model is applicable.Both CET-4 and CET-6 are the important index of the measure of the college English teaching achievement and undergraduates' knowledge degree about English. These tests are authoritative, strict and comparable. By stratified cluster random sampling, this study chose all of full-time undergraduate in Shandong University in Grade 2003 as the objects of study, and collected some information of each student. Apply univariate analysis, generalized linear regression models, logistic regression models and multilevel models to analyze examination results and influential factors of CET-4 and CET-6. Use different multilevel model based on types of dependent variables, and compare the result of multilevel model to generalized linear regression model and logistic regression model. This study aimed to appraise and explain the influence of the college English test, to provide a scientific basis for improving the English level in our university, and to evaluate the applicability of multilevel model on examination results. Results:1. Examination rate of CET-4 is 75.4%, mean and standard deviation were 65.84 and 13.10, and pass rate of students who took test, excellence rate and pass rate of all were 69.9%, 6.0% and 52.7%. The indexes of CET-6 are 40.7%, 67.83, 8.48, 38.2%, 93.8% and 8.3%.2. Univariate analysis:â‘ The influential factors of scores and relative rates of CET-4: gender, age, nation, educational system, place of birth, semester and discipline. Among them, the number of girls who took CET was less than boys, but their level of English is higher than boys; Younger students' English level is higher; Han students have higher level of English; The English level of students with long-system is relatively higher; Students from Shandong Province is better than students from other provinces; The English level of students who took test earlier is higher, and also less difference; Art students have low level of English.â‘¡The influential factors of CET-6 were gender, age, nation, educational system, place of birth, semester and discipline. Among them, the girls' test results and pass rate of students, excellent rate and examination rate is higher than boys; The influence of age, nation, education system, birthplace and semester is as the same as CET-4; Medical Students have the lower level of English.3. Multivariate analysis :â‘ Generalized linear regression models showed that the influential factors of CET-4 scores are gender, age, nation, educational system, birthplace, semester and discipline; the influence of CET-6 are gender, educational system, birthplace and semester.â‘¡Logistic regression models showed that the influence of the CET-4 relative rates are mostly the same as CET-6, and the main of them are gender, age, educational system and discipline.â‘¢Multilevel Model for numerical data showed that the influential factors of CET-4 scores are gender, age, nation, birthplace, semester and discipline; The influential factors of CET-6 are gender, age, educational system, birthplace, semester and the CET-4 passing rate of school.â‘£Multilevel Model for binary data showed that the influence of the CET-4 relative rates are mostly the same as CET-6, and the main factors are gender, nation, birthplace and academic types. Conclusion: While multilevel model was used to analyze the scores and relative rates of the college English test, its result is better than the result of traditional methods of regression analysis (generalized linear regression model and logistic regression model).
|
PDF Full Text Request