Analysing Student Evaluations of Teaching: comparing means and proportions

Student Evaluations of Teaching (SETs) play a central role in modern academia. They are used for tenure, promotion, teaching improvement and other important decisions. One would think that the data collected from a SET would be analysed correctly, but such is typically not the case, as can be seen in this study later. Therefore we propose a correct method for analysing SET data. The present paper compares the two methods on a large data-set of actual SETs. We show that the traditional method can misrepresent a teacher's performance, and that the traditional method can be extremely sensitive to outliers; neither of these characteristics is desirable. In contrast, the proposed method appears to suffer from neither of these defects.