4 ms·

This is one of those complex topics which on HN, seesm to lead to "it's simple, do X" types of analyses, which tend to be simple, confident, and wrong. Some fa

by xmddmx 4y ago

This is one of those complex topics which on HN, seesm to lead to "it's simple, do X" types of analyses, which tend to be simple, confident, and wrong. Some facts to consider: At Harvard, over 40% of non-minority students received special treatment for admissions (which includes factors such as being a legacy, athlete, or the 'dean's list' which was a list of high donating parents). 75% of those would not have been admitted w/o that help. I think this lawsuit is trying to pierce this veil. [1] For those that argue "meritocracy" and that standardized testing is the answer: the SAT has a correlation of about r=0.3 to 0.5 with first year college GPA. Thus, the explained variance (r*2) is about 9% to 25%, meaning as much as 81% of college success is not statstically predicted by the SAT. [2] [1] https://www.theguardian.com/commentisfree/2021/nov/17/harvard-university-students-smart-iq https://www.theguardian.com/commentisfree/2021/nov/17/harvar... [2] https://satsuite.collegeboard.org/media/pdf/national-sat-validity-study-overview-admissions-enrollment-leaders.pdf https://satsuite.collegeboard.org/media/pdf/national-sat-val...

fsckboy 4y ago
> 75% of those would not have been admitted w/o that help this not making a claim anywhere near as interesting as it pretends. Since you're using a %, what is the % difference in quality that admitting these 75% makes (assuming it's even true, considering I don't believe much from the Guardian) and how does that (what the author calls "affirmative action") compare with other affirmative actions taken to achieve a balanced class? were they otherwise 99% good enough, or only 10% good enough? My money is on above 90% because I went to an elite school (that has no legacy) and lots of people there were extremely highly qualified, and the ones who weren't did not fall in a pattern of being wealthy or whatever, probably better explained by something like your evidence of how ineffective IQ (what SAT does correlate to) or high school grades are in predicting college GPA and how effective GPA is or isn't in predicting future contributions. I find grades have much more to do with psychological makeup, like do you rebel against authority, or knuckle under. Good grades are valuable, but don't tell the whole story of your value as a contributor, sometimes screwed up and crazy genius works better. But either way, shouldn't that be up to the school what combinations they wish to place their bets on?
- charlieyu1 4y ago
  Grades are the only metric that is difficult to cheat even with wealth.
  - Jcowell 4y ago
    Couldn’t you just bribe people?
    - naniwaduni 4y ago
      Not all bribery is equally easy.
  - fsckboy 4y ago
    unless you have especially rigorous grading (and look at the volume of comments here bitching about grading on a curve, so it's rare), it's easy to get good grades by just doing the normal amount of work, put in some time each day on each class. It's a measure of conscientiousness, not smartness, ability, etc., but yes, it's a good measure of getting a baseline amount of learning out of a course, nothing exceptional. (people here who bitch about grading on a curve should take a course in statistics graded on a curve, then listen to their fellow students who got A's to explain how it works.)
  - washadjeffmad 4y ago
    The private "school of excellence" I attended instituted a 40% minimum on any assignment turned in with a name and date on it, and labs, tests, and quizzes - combined - made up 25% of one's final grade. The administration contended this policy had nothing to do with the $8M athletics facility built with donor funds, nor the influx of scholarships awarded to students who all happened to join the football team that year. I disagreed. Vocally. When I went from a 98% student to a 24% student at the start of a new semester just by refusing to write my name on any homework, I was pulled from all activities, isolated in endless detentions, and put on a "remedial track" away from my class (like ISS). It took a few more weeks of talking with the teachers and priests and protesting, but the headmaster eventually stepped down and was replaced by one of the nuns. Grades aren't hard to cheat if the institution is complicit, especially if what's being measured is compliance and obedience.
yodsanklai 4y ago
> For those that argue "meritocracy" and that standardized testing is the answer: the SAT ... Maybe other standardized tests are or could be better predictors. Some countries rely exclusively on them for admission (and if I'm not wrong, also the US is some specialized fields such as law or medicine).
- xmddmx 4y ago
  The SAT is one of the mostly highly researched tests. I'm curious of your claim: Please, name one, and cite a source.
Tozen 4y ago
Seems like you are going down the path of presenting the reality, where a lot of people will ignore "wealth favoritism", cronyism, nepotism, etc... But mention affirmative action, specifically minorities of color as oppose to women of a certain persuasion (also ignored or given a pass), and things can get real hot and tribalistic, real fast. Well sir/madam, good luck with your arguments and evidence. At the very least, as it's doubtful such biases or prejudices that protects systematic advantages and socioeconomic position will be altered, it will make for a good popcorn read.
credit_guy 4y ago
It does not look to me the SAT study you linked to concludes what you claim it concludes. You are saying in 81% of the cases the SAT score is useless. They are saying this: When HSGPA (high school GPA) and SAT are combined, the correlation with FYGPA jumps to .61, an increase of .08 and a 15% boost in predictive utility over using HSGPA alone.
- xmddmx 4y ago
  I was replying to another comment which suggested using only standardized testing. It's true, when you combine SAT with GPA, it's better, but that was not the point I was rebutting. Even so, with GPA+SAT, if r = .61 then r^^2 = .37, meaning 63% of variation in college grades are not explained. And I believe this is first year college GPA only - "getting a good GPA in freshman year" is probably a weaker argument than "succeding in college, or afterwards"
  - credit_guy 4y ago
    I never heard of anyone suggesting only SAT. What I heard instead (and what's happening in real life) is universities ditching the SAT. Countless universities did that (e.g. those in the U. California system). The stated reason is equity, but it's often followed by "it's redundant anyway". Well, it's not redundant. And this study shows that. Now you are saying that the SAT does not explain much of the freshmen year GPA. But that could be (and it probably is) a consequence of the fact that colleges have different standards. The SAT could have 100% prediction power, like the Oracle of Delphi, if the colleges have different standards the correlation between SAT and GPA will be low. A second effect is at play here, the binning effect. If you have two variables X and Y that are highly correlated (let's say rho = 99%),and then you split the observations in bins based on X, then the correlation of X and Y for each can be very low. They could be arbitrarily close to zero. Just give it a try. Finally, prediction power beyond the first year. Why would colleges need that? They need quantitative metrics to determine who would perform well once they come in. After that, it's their duty to educate and nurture those students. If they can't do that, and the final outcome is some form of lottery ticket, that's on them, not on the SAT.

by xmddmx 4y ago

fsckboy 4y ago

> 75% of those would not have been admitted w/o that help this not making a claim anywhere near as interesting as it pretends. Since you're using a %, what is the % difference in quality that admitting these 75% makes (assuming it's even true, considering I don't believe much from the Guardian) and how does that (what the author calls "affirmative action") compare with other affirmative actions taken to achieve a balanced class? were they otherwise 99% good enough, or only 10% good enough? My money is on above 90% because I went to an elite school (that has no legacy) and lots of people there were extremely highly qualified, and the ones who weren't did not fall in a pattern of being wealthy or whatever, probably better explained by something like your evidence of how ineffective IQ (what SAT does correlate to) or high school grades are in predicting college GPA and how effective GPA is or isn't in predicting future contributions. I find grades have much more to do with psychological makeup, like do you rebel against authority, or knuckle under. Good grades are valuable, but don't tell the whole story of your value as a contributor, sometimes screwed up and crazy genius works better. But either way, shouldn't that be up to the school what combinations they wish to place their bets on?

charlieyu1 4y ago
Grades are the only metric that is difficult to cheat even with wealth.
- Jcowell 4y ago
  Couldn’t you just bribe people?
  - naniwaduni 4y ago
    Not all bribery is equally easy.
- fsckboy 4y ago
  unless you have especially rigorous grading (and look at the volume of comments here bitching about grading on a curve, so it's rare), it's easy to get good grades by just doing the normal amount of work, put in some time each day on each class. It's a measure of conscientiousness, not smartness, ability, etc., but yes, it's a good measure of getting a baseline amount of learning out of a course, nothing exceptional. (people here who bitch about grading on a curve should take a course in statistics graded on a curve, then listen to their fellow students who got A's to explain how it works.)
- washadjeffmad 4y ago
  The private "school of excellence" I attended instituted a 40% minimum on any assignment turned in with a name and date on it, and labs, tests, and quizzes - combined - made up 25% of one's final grade. The administration contended this policy had nothing to do with the $8M athletics facility built with donor funds, nor the influx of scholarships awarded to students who all happened to join the football team that year. I disagreed. Vocally. When I went from a 98% student to a 24% student at the start of a new semester just by refusing to write my name on any homework, I was pulled from all activities, isolated in endless detentions, and put on a "remedial track" away from my class (like ISS). It took a few more weeks of talking with the teachers and priests and protesting, but the headmaster eventually stepped down and was replaced by one of the nuns. Grades aren't hard to cheat if the institution is complicit, especially if what's being measured is compliance and obedience.

yodsanklai 4y ago

> For those that argue "meritocracy" and that standardized testing is the answer: the SAT ... Maybe other standardized tests are or could be better predictors. Some countries rely exclusively on them for admission (and if I'm not wrong, also the US is some specialized fields such as law or medicine).

xmddmx 4y ago
The SAT is one of the mostly highly researched tests. I'm curious of your claim: Please, name one, and cite a source.

Tozen 4y ago

Seems like you are going down the path of presenting the reality, where a lot of people will ignore "wealth favoritism", cronyism, nepotism, etc... But mention affirmative action, specifically minorities of color as oppose to women of a certain persuasion (also ignored or given a pass), and things can get real hot and tribalistic, real fast. Well sir/madam, good luck with your arguments and evidence. At the very least, as it's doubtful such biases or prejudices that protects systematic advantages and socioeconomic position will be altered, it will make for a good popcorn read.

credit_guy 4y ago

It does not look to me the SAT study you linked to concludes what you claim it concludes. You are saying in 81% of the cases the SAT score is useless. They are saying this: When HSGPA (high school GPA) and SAT are combined, the correlation with FYGPA jumps to .61, an increase of .08 and a 15% boost in predictive utility over using HSGPA alone.

xmddmx 4y ago
I was replying to another comment which suggested using only standardized testing. It's true, when you combine SAT with GPA, it's better, but that was not the point I was rebutting. Even so, with GPA+SAT, if r = .61 then r^^2 = .37, meaning 63% of variation in college grades are not explained. And I believe this is first year college GPA only - "getting a good GPA in freshman year" is probably a weaker argument than "succeding in college, or afterwards"
- credit_guy 4y ago
  I never heard of anyone suggesting only SAT. What I heard instead (and what's happening in real life) is universities ditching the SAT. Countless universities did that (e.g. those in the U. California system). The stated reason is equity, but it's often followed by "it's redundant anyway". Well, it's not redundant. And this study shows that. Now you are saying that the SAT does not explain much of the freshmen year GPA. But that could be (and it probably is) a consequence of the fact that colleges have different standards. The SAT could have 100% prediction power, like the Oracle of Delphi, if the colleges have different standards the correlation between SAT and GPA will be low. A second effect is at play here, the binning effect. If you have two variables X and Y that are highly correlated (let's say rho = 99%),and then you split the observations in bins based on X, then the correlation of X and Y for each can be very low. They could be arbitrarily close to zero. Just give it a try. Finally, prediction power beyond the first year. Why would colleges need that? They need quantitative metrics to determine who would perform well once they come in. After that, it's their duty to educate and nurture those students. If they can't do that, and the final outcome is some form of lottery ticket, that's on them, not on the SAT.