中文摘要：

以概化理论为基础，探究影响高校教师教学水平评价结果的因素。采用《高校教师教学水平评价量表（学生用）》收集评价数据，用mGENOVA对数据进行分析。结果发现：（1）与在第一学期末进行教学水平评价相比，在第二学期初进行教学水平评价的结果可靠性更高；（2）评价每位教师的教学水平仅需抽查20名学生即可保证评价结果的可靠性；（3）不同专业类型的学生对评价指标的侧重点不同，继而影响评价结果的可靠性；（4）学生对理科课程的评价可靠性较高，对文科课程的评价可靠性较低。

英文摘要：

In order to improve the quality of college education, colleges in China have an evaluation for college teachers＇ teaching every semester. Generally speaking, colleges have students evaluate each one of their curriculum teachers. Since each teacher teaches a different class, there will be few results for each of them, and a concordant conclusion for each teacher cannot be drawn. Furthermore, there are multiple factors which influence results. For example, too few students taking part in the evaluation leads to low reliability of the results, students with different majors pay attention at different levels to indexes and different kinds of curriculums earn disproportionate grades. Because of students focusing on different issues through different periods, point in time is another factor that affects results. This study, based on the generalizability theory, offers a method to solve the problem mentioned above and discusses factors that affect teaching evaluation results for college teachers. The data was collected by the scale of Teachers＇ Teaching Evaluation and collected from 19 curriculums, 7 liberal arts curriculums and other 11 science curriculums. 558 data points were collected at the beginning of the semester on March and 566 collected at the end of the semester in December, of which involved 5 liberal arts majors, 10 science majors and 4 engineering majors. All data was saved to the txt format and analyzed with mGENOVA. According to the generalizability theory, evaluations taken by a specific number of students is reliable enough to measure one teacher. The generalizability theory uses an index of dependability （φ） instead of validity used in classical test theory to judge the reliability of results. In terms of the needed number of evaluators, the D study result shows that reliability increases while the number increases, and it is appropriate to have 20 students evaluate each teacher. The study also finds out that students majoring in engineering courses who pay more attention to practical is