Teaching is a multidimensional process comprising a number of aspects, e.g., instructor attributes, which sometimes are difficult to evaluate. In particular teaching effectiveness, that is an aspect of teaching, is influenced by a combination of teacher characteristics (such as clarity, capacity to motivate the students and to help them in the study of his topic, ability to organize the lesson also with exercises and handouts, for example but also gender, age, previous experiences), physical aspects of the classroom or laboratory (too crowded or with an insufficient number of computers) and class characteristics (such as students' characteristics: gender, age, high-school of origin, mark obtained at the end of compulsory or high school, faculty attended by the student, or class size). As teaching effectiveness is becoming even more important in a system of school evaluation (in our case university system of evaluation), it is necessary to find how to measure it. This paper considers the problem of assessment of teaching effectiveness from students' point of view, analysing the questionnaires given to the students of the University of Udine at the end of their courses. The problem, in statistical terms is to relate an 'outcome' variable (the dependent one), in this case ratings given by the students or a particular linear combination of it, to a set of explanatory variables both to the student and teacher level. The data set used in the analysis consists of almost 9500 questionnaires regarding 416 courses of the University of Udine covering the academic year 1999-2000, its structure (questionnaires clustered in courses) suggest the use of a particular class of regression models, the multilevel models.