<p>
</p>
<p>Well, actually, the better answer seems to be that the dataset suffers from the pervasively invidious problem of statistical selection bias due to truncation; it doesn’t record the results of the students who drop the class, probably because they know they will receive poor grades. Those who have taken the lower-division weeders will surely remember that the number of students enrolled at the end of the course will be substantially smaller than the number of students enrolled at the beginning, as many students will have jumped, particularly after viewing their grades from the first midterm. What happens to those students? They should be evidenced by either a large bump of ‘F’ grades denoting those students who would have failed had they stayed in the class (because they got scores of zero on every assignment and exam after they had dropped, hence surely failing had they stayed in the course), or by a specific category of ‘drops’. The data shows neither, indicating that selection bias is likely rampant.</p>