undefined | Better HN

0 pointszarify15y ago0 comments

It's hard to argue with the logic of that approach. Classrooms aren't always logical places though. Consider factors like student movement between classes, courses and programs designed to move students to the workforce rather than complete school, parental pressure on students and teachers to place students into unrealistic pathways, and so on.

I suppose you could factor all that into the model, though I'd question the feasibility of doing so.

I'd also argue that grades aren't the only important factor in schooling. Building social skills, discipline, time management, helping to realize goals... All ways in which good teachers can influence students, and only some of them reflect in neat packages like test scores.

Edit: I'd like to clarify that in my previous post I mentioned getting more difficult classes sometimes as the luck of the draw. I didn't mean low ability classes. I meant classes with students with serious behavioral issues, students with learning disabilities and so on. These will very often negatively impact the learning outcomes of the entire class, and is something you very rarely if ever have to deal with in top end classes.

0 comments

1 comments · 1 top-level

yummyfajitas15y ago

Consider factors like student movement between classes, courses and programs designed to move students to the workforce rather than complete school, parental pressure on students and teachers to place students into unrealistic pathways, and so on.

This would only matter in the event it did not average out over a teacher's entire cohort of students. I.e., you would need to get a situation where all the students who wanted to move into the workforce were taught by teacher A, and all the students who wanted to complete school were taught by teacher B.

The law of large numbers helps here.

Also, this would only matter insofar as this factor is uncorrelated with the things you do use as predictors - parental income, race, prior grades, etc. If all the low income students want a job, and all the high income students want to complete school, the predictor already incorporates this factor.

I meant classes with students with serious behavioral issues, students with learning disabilities and so on. These will very often negatively impact the learning outcomes of the entire class, and is something you very rarely if ever have to deal with in top end classes.

If this factor is strongly correlated with income (as you suggest), the predictor will already incorporate it.