Tuesday, May 15, 2012

Grading, Correlation vs. Causation

Final grades are finally in, thank goodnees.  With 100+ students, even entering the grades is non-trivial. 

Once again, freshman in my undergraduate Algorithms and Data Structures class receive A's at a substantially higher rate than other students.  (Way, way higher.)  My conclusion:  it's best for your grade if you take my class as a freshman.

13 comments:

Anonymous said...

Of course, my students are smarter.

It's the other professors who are inflating grades unnecessarily.

Anonymous said...

Harvard grade inflation is very well known, and it seems Michael is not helping in these matters.

Michael Mitzenmacher said...

I'm confused. How are you all getting grade inflation out of this? All I'm saying is that, of the As received in my class, a surprising number of them go to freshmen. I don't grade freshmen differently, and I haven't (and wouldn't) discuss the number of A grades on the blog.

Anonymous said...

Wow...this isn't youtube; please keep ignorant and obnoxious comments to yourself. Some of us appreciate an inside look at teaching classes and don't want others dissuading the bloggers.

Wim van Dam said...

Interesting. Without revealing sensitive information, can you quantify "Way, way higher"? Do you have any idea why they do better than sophomores, etc?

- Wim

Anonymous said...

in fairness to the first two commenter(s) on first reading I also parsed the post as saying Michael gives a lot of A's in his class compared to other classes. A second reading revealed he was simply saying freshman tend to do better than upperclassmen in his course. And then a joke about an individual student taking the class sooner and doing better :)

Anonymous said...

I think Anons #1 and 2 are interpreting "receive A's at a substantially higher rate than other students" as "higher rate than other students in other classes".

JeffE said...

I've noticed the same effect. Students who take my undergrad algorithms class way ahead of schedule are much more likely to earn As or A+s. I don't remember any freshmen in my class — most students are seniors or late juniors — but I do get a few first-semester sophomores.

But there may be some selection bias at work here. If you're taking a class two years early, and it isn't going extremely well, why _not_ drop and retake it later?

Anonymous said...

Former undergrad here.

Just for the record, Professor Mitzenmacher's algorithms class is generally known to be a difficult class to get a good grade in. Whatever your feelings about Harvard and grade inflation are, CS 124 is really not a good example.

Anonymous said...

I'm an upperclassmen who knows the various freshmen taking the class this year. The joke that people should take 124 as freshmen, is, of course, a joke, but I guess, just to explain for those who aren't form these parts:

The freshmen who are likely to take 124 are pretty much ahead of schedule in terms of their grasp of mathematical rigor and problem solving (not to put down their achievements at all). In fact, quite the opposite -- they're brilliant and work extremely hard for it, more power to them. We have IMO and USAPHO students among them this year, along with a bunch of Math 55ers.

Interestingly, this year's 124 freshmen are super helpful, sociable, and nice, too, something I've found uncharacteristic for past years of 55/IMO/USAPHO. I heard they even took Prof. Mitzenmacher out to lunch some days in Annenberg. I don't think that's like, ever happened.

Arjun Narayan said...

@Jeff: I think it's a different kind of selection bias. What kind of freshman takes a hard class ahead of schedule? The kind of freshman who is just way too good to get a bad grade regardless of when he/she takes the class.

The rest of the students who take it at the average time to take it are probably... average* and get average grades (i.e. distributed around the non-freshman mean).

You're selecting out the geniuses (by the instrumental variable of taking classes early), and this genius population is going to have a qualitatively different set of outcomes.

* Although "average" at Harvard means something completely different from the usual sense of the word...

Michael Mitzenmacher said...

Comments on comments:

Just for the record, Professor Mitzenmacher's algorithms class is generally known to be a difficult class to get a good grade in. Whatever your feelings about Harvard and grade inflation are, CS 124 is really not a good example.

I've heard that too!

I admit I still don't get how one reads what I wrote and reads it as I'm giving out a lot of As compared to other classes, but thanks to all for clearing it up.

I purposely avoided giving my "interpretation" to the data, but since it's come up...

Jeff -- I think the number of drops/retakes I've seen while at Harvard for my class can be counted on 1 hand. (Maybe, even, 1 finger or less.) There's no culture here of "re-taking" a course for a higher grade that I know of.

Nomad I think has it right -- the freshmen who take my course are generally very, very strong.

And just to clear it up, I make myself available to have lunch with all students. The fact that I lunched with freshmen, who were indeed very nice, was not the reason for their good grades.

DoubleRobustEstimator said...

@Nomad has the right answer, I think. I believe, but can't locate a reliable reference, that in tests like the AMC 10, "younger grade" students tend to have much better averages/median scores, etc. than the "grade appropriate" students, for essentially the same reason.