Keep up the good work –

Steven: I’ve got to think doing this assignment would bring about a whole bunch of odd feelings in both the students, the customers, and the cashiers.

Certainly. So we talked for several minutes about ways to mitigate that weirdness, all the way from making sure you introduce yourself to the cashier to covertly recording the transaction with your cell phone’s video camera and analyzing the data later.

Still, I won’t say I was surprised by the low submission count.

I like Ben‘s note that students should be able to “mime” out the approximate times and I like how Kevin takes it farther by having the students set up roles as checker and shopper and gather data on themselves. I imagine there’s a large degree of uncertainty here, where students who know they’re being timed will work faster but it couldn’t have hurt to model exactly the kind of data collection I was assigning more explicitly.

The poorly faked data points to the fact that students don’t have that vocabulary. Students may not be as intimately familiar with the check-out process as us weekly grocery shoppers, but they know the drill. They should know the drill well enough to be able to “mime” out how long checking out 10 or 100 items takes (à la Dan the Record Breaker). Perhaps the value of the faked data isn’t so useful as an opening to outliers or linear regression- perhaps it’s a more valuable as an authentic assessment of students’ mathematical reasoning ability.

Personally, I disagree with Jane and think you should tackle the topic of which are outliers. While statistically speaking it’s a tricky and complicated business, there is value in looking at data and being able to determine where something went wrong.

I’d also bash ’em for faking data period. Not cool.

What to do about them is NOT a subject for anyone who has had less than two or three college level statistics or econometrics classes.

If you are going to teach linear regression…I’d suggest getting a well behaved data set and reuse it year after year. That way, you can focus on the lesson and you don’t have to deal with the data anomalies.

FYIW, 32 data is barely enough for a central limit theorem to kick in, assuming independence and normality. If the data points are not collected from different checkers, the independence assumption may not hold.

Yes, I have a Ph.d with a field in econometrics.

While I love the lesson and the assignment, I’m wondering if the situation has too many stressors in it?

I may try to have it both ways, though, and discuss the above graph in class this week.

Strong stuff.