Hacker News Reader

Comment by BoorishBears | Hacker News Reader

"Mean SAT score for non-Jewish Whites was derived by assuming the mean Jewish SAT score is a generous 1268."

Very scientific.

And the conclusion:

"Their 100- and 177-point SAT score lead over Hispanic and Black students seems to have hurt, not helped, their chances of admission.

On the other hand, the smaller 73-point lead of Asian students over Whites helped greatly."

They're using scores from across the 100% of college applicants when less than 1% of them apply to Ivy League schools.

If they had used more natural language I'd be charitable and say they didn't know better, but not after they went out of their way to invent numbers to... separate out Jewish people from whites? While also insisting they're a small group so it doesn't matter?

I'm going to guess the admissions numbers didn't show the kind of dramatic gap they wanted so off to town they went.

View on HN

> While also insisting they're a small group so it doesn't matter?

Doesn't matter for calculating the non-Jewish White SAT is how I interpreted that. Since white_SAT = (2.4*jewish_SAT + 55.4*non_jewish_white_SAT) / (2.4 + 55.4), it means that even if your Jewish SAT estimate is off by ~100 points, it will only change the non-Jewish White SAT estimate by ~4 points.

Is there something non-scientific about estimates? Or do you think this is not a reasonable estimate, i.e. it may be significantly off? The charge of "non-scientific" is very vague - can you be more specific? What do you think is wrong, and by how much?

Specific to the estimate... it's not scientific to guess a number because you proved guessing isn't a risk because it's such a non-factor. In general you just remove the non-factor.

But, realizing you're OP...

> They're using scores from across the 100% of college applicants when less than 1% of them apply to Ivy League schools.

Respectfully, you need to go back to high school stats if you're trying to salvage this. Or maybe a bit earlier, since I'm pretty sure my 8 year old niece could see what's wrong there.

> it's not scientific to guess a number because you proved guessing isn't a risk because it's such a non-factor. In general you just remove the non-factor.

But then you don't have a reasonable lower bound, and someone trying to debunk the analysis could, instead of throwing un-falsifiable claims of un-scientificness, make the much stronger accusation that the combined white+Jewish mean SAT score is being unfairly attributed to non-Jewish whites alone.

> They're using scores from across the 100% of college applicants when less than 1% of them apply to Ivy League schools.

Seems relevant to me - it would be very surprising if the mean of the SAT score distribution didn't strongly affect its top percentile. If you're not satisfied with reasonable extrapolation, you can find what the SAT scores are for the top percentile of white/Hispanic/Black students. I wager that sorting the SAT scores by mean or by top percentile gives the same order. But as you say, I failed high school stats, so I'm probably wrong.

Reply on news.ycombinator.com