Physics students, four subsamples.

One of the important choices in analyzing the data was to analyze four groups of respondents separately: male regular physics students (MR), female regular physics students (FR), male advanced physics students (MA) and female advanced physics students (FA). In the total population of Dutch high school physics students, regular physics students outnumber advanced physics students and males outnumber females. The fact that advanced physics students and especially females are overrepresented in my sample compared to the total population has the advantage that the four groups of students are quite balanced in numbers. Only for the group of MR-students it should be taken into account that the number of respondents is on the verge of being insufficient for statistical analysis, not only because of the relatively low number of respondents but also because of its low representativeness.

One of the most important arguments in analyzing regular physics and advanced physics separately is the fact that the final exam for advanced physics is hardly comparable to the final exam for regular physics.  In analyzing regular and advanced physics students together differences in the exams could easily obscure the influence in other parameters that are of more interest to our research objective.

From the literature, but also from preliminary analyses, I concluded that gender differences in many parameters, e.g. quantity and quality of work, turn out to be substantial. As Hazari , Tai, and Sadler (2007) already pointed out, there are 'pedagogies and affective factors that might influence male and female students differently'. In an analysis of all the respondents together, gender differences are apt to obscure the potentially more subtle relationships between student and teacher. By separating the sample in different gender groups, I was able to detect these kinds of subtle relationships with straightforward statistical techniques which, in the specific circumstances of this study, facilitated a reliable interpretation of the data.