Say that, out of 180 responses in an online survey sent out during a week in March, 60 respondents indicated that they took over forty-five minutes to get to university each way. Of these students, 35 per cent said that they’d been late to or missed at least one morning class in the last month, and of the 120 remaining respondents 20 per cent indicated that they’d done the same. It’s plausible that students with longer journeys to university miss more classes, but it’s also possible that some students who answered this survey question conscientiously and in good faith have overestimated their typical journey time (maybe their journey was unusually long on the day they filled in the survey), or underestimated how often they’ve missed morning classes (maybe they’re aware of the cases in which they struggled to get to morning classes on time but forget the many in which they got there without too much of a fuss). Perhaps there’s some other factor that affects both journey times and attendance, namely employment, caring responsibilities, disability, the way timetables are structured, whether or not there are online resources that mean students can still keep up if they miss classes, and students with longer journeys to university are also more likely to have one or more of these.
This hypothetical survey has other limitations too, of course. Only 180 people responded, which might not be enough to generalise beyond the sample. The responses might not be a representative sample of all students at that university, perhaps because the survey was only circulated on two email mailing lists and only students who cared about journey times to university responded, or maybe the people with the longest journeys to university and the least free time didn’t get round to filling it in, and certainly not all students at all universities. Even if we increased the number of responses to the survey we couldn’t necessarily get rid of this potential sampling bias.
If we wanted to find out more we might survey students at a range of universities, ask more questions (about the sorts of potentially confounding factors mentioned above) and follow up with students after a long period of time to see if there were any connections between journey times and lateness/missed classes. If we were particularly interested in timetabling we might look at lateness/missed classes rates across timetables and types of student and try to account for the reasons why students sign up for certain classes and not others.
Until then, though, we can usefully discuss this hypothetical survey and its results in class and raise questions about survey design, confounding factors, sampling bias and so on. We might also usefully discuss it if we’re trying to work out how we should structure timetables, not to jump to conclusions and say, for instance, that making everyone’s journey to university less than forty-five minutes would have some specific beneficial effect on lateness/missed classes rates, because we can’t infer that from this data, but as a basis for further investigation and discussion.