<p>You’ve got it backwards. It isn’t the perfect SAT scores that cause the kids to be high achieving. The high achieving kids get the perfect SAT scores. To a degree high SAT scores are predictive of who will be admitted to the top schools but they do not cause the applicants to be admitted, a fine distinction, but I think it is a distinction some of us are trying to make.</p>
<p>Let’s say I had an admittedly strange assignment and I was supposed to predict whether or not someone would be admitted to a top school. In this strange scenario I have a choice, I can either look at their test scores or I can meet them for about an hour during which I’m not allowed to ask about grades or test scores.</p>
<p>I’d take the interview. I would try to ascertain how smart they were and whether or not they seemed to be one of the high achievers who get into these kind of schools. I think that would give me more information about the outcome than raw test scores.</p>
<p>Because it isn’t the scores that drive the decisions, it is the type of applicant.</p>
<p>If you try and define more precisely what you mean by terms like “high achieving” and “cause”, you will find that perfect SATs are essentially causal here in a way that the other indices of performance are not. The SAT does not physically cause anything. But if you try to measure whatever it is that does cause both the perfect SAT and the high other factors, it will correlate far better with SAT than with the other standard performance metrics. One indication of this is that having high SAT tends to imply having high other metrics but not the other way around. People with a determination to achieve are in much more common supply than extremely high SAT scorers; if valedictorians had the ability to generate perfect test scores by strategically studying or preparing or praying, there would be a lot more of those 2400s.</p>
<p>The exception to this is if you have information about even more selective metrics in addition to perfect SAT, such as winning a national math olympiad. These are (effectively) causal of perfect SAT in the same way that perfect SAT is effectively causal of the high grades or impressive schoolwork. In such a situation you would find that winning the national olympiad is “causal” of everything else, because it is far and away the most revealing measure (among those used) of the latent and otherwise hard to observe traits that one is really looking for.</p>
<p>Do you have actual data for this? I’d be interested to see it. </p>
<p>I tutored for Ivy West for several years and saw many people, including myself, improve their test scores through practice and technique, without any other corresponding improvement in other areas, or change in base intelligence.</p>
<p>
This comment is of course, right on the money. It depends on what you mean by “high achieving.” For example, I don’t know that there is necessarily any correlation between SAT scores and future income level, or that the correlation is stronger than with any other metric. Although, like I said, I’d be interested to see actual data.</p>
<p>To improve to 2400 you need to start out high. You do not, in general go from all 550’s to all 800’s.</p>
<p>With respect to data, Harvard did internal statistical studies indicating that extremely high SAT scores correlated well with magna and summa cum laude degrees. Presumably they had other variables in the regression, but SAT was the one commented upon after the results were in.</p>
<p>High achieving was covered early on in the thread. It took me awhile to get my head wrapped around, it is essentially the kids who operate at a high level. They have the high grades, high test scores and impressive extra-curriculars and they make it look pretty effortless. I can’t define it any better than that, well, I could try but I don’t think it would convey very well. I’ve known kids like this and they were accepted everywhere they applied.</p>
<p>I agree, the SAT results correlates with an applicant being a high achiever, as do other performance metrics. But a predictive model would only interest me if it got to the root of why the kids who are accepted are accepted. I think the reason cannot be captured by standard performance metrics.</p>
<p>Presumably, the study was based on Harvard student data?</p>
<p>Since there is a limit on summa and magna degrees, it must mean that Harvard admitted students who did not have extremely high SAT scores. Granted, some may have been hooks (though I would like to see a study of how hooked vs. non-hooked students performed in college).
This gets us back to the issue of Ivy-caliber students and predictions of their admission into the Ivies+: Harvard and other Ivies+ admit students with less than extremely high SAT scores.</p>
<p>So, guess how many students nationwide get “extremely high” SAT scores at a single sitting (e.g., 2300s or above?)</p>
<p>This discussion about super high SAT scores is so far into the long tail of the population of potential Ivy admits that it deserves its own thread, entitled “Extremely High SAT Scores Enhance Admissions Chances at Top Schools” :-)</p>
<p>Kei</p>
<p>Answer: about 6,000, ~ half the frosh slots at Ivies.</p>
<p>And … if 6,000 kids have the “extremely high” scores, and that is about half the frosh slots at Ivies, it also doesn’t follow that they actually take up half the frosh slots at Ivies. Many, many kids with extremely high scores go to state flagships, because their state flagships are good enough, they want to stay in-state, and / or the price is right.</p>
<p>PG, what is it in this topic that got you so upset? </p>
<p>Where these 6000 kids went to college is not the focus of the topic. Whether they got admission into one of the top school or not is. I think teh idea is that a much higher % of students with these “extremely high” scores do get into one of the top schools or they are IVY caliber. Where they go is a complete different story.</p>
<p>I believe the point that PG and Kei-O-lei are trying to make is that at least half of Ivy admits do not have extremely high scores. So this is one metric in the formula proposed by PCP that works for fewer than half of Ivy admits (and even less is we consider Ivy+ schools).</p>
<p>Some schools I think it is easier to make a fairly predictive quantitiative model, particularly some of the UCs (like Berkeley) where they use precisely defind “Comprehensive Review.” </p>
<p>Lots of nuggets of wisdom from earlier posts. </p>
<p>
</p>
<p>I agree with the significance of this factor. How do we capture this “interesting uniqueness”? </p>
<p>
</p>
<p>It is definitely more impressive when an achievement is accomplished in hardships. Adcoms do take context in consideration, but far from enough in my opinion. If everything is compared in or adjusted for context, then it is unlikely we would have half full-pay students at Harvard. It is hard for mortals not to fall for spectacular absolutes. How can we factor in context?</p>
<p>
</p>
<p>I agree with these tradeoffs, and this is why the model has weightings. Granted, it is a challenge to integrate these types of tradeoffs.</p>
<p>
</p>
<p>I wholeheartedly agree with the second point. How can we capture the “noncompulsory” factors? It is easy to spot if it is part of school curriculum, but how do you separate truly noncompulsory achievements from parent-pushed ones? Does the difference matter?</p>
<p>Regarding the first point, the argument presented is not about the merit of 2400 versus 2300. By themselves, 2400 is better than 2300, however slightly.</p>
<p>
</p>
<p>It is hard to capture the this one. Very few kids get the opportunity to showcase their ability live to the school of their choice, but if they get the chance, I agree it can be decisive.</p>
<p>
</p>
<p>This is may be true in general, but the priority is definitely reversed in magnate or highly ranked high schools. Being a val of Stuyvesant or TJ is much more impressive than a 2400.</p>
<p>
</p>
<p>I’m inclined to agree with this. CDS’s from Ivies+ (I’ll stick with this term for now) show much higher acceptance rate for extremely high SAT scorers than the acceptance rate for other tracked stats.</p>
<p>
</p>
<p>Even if we build the model with this kind incremental approach, little things add up! 5% here, 10% there, pretty soon you have tripled your chances.</p>
<p>
</p>
<p>You said it. The point IS to have high prediction success. I would consider a huge success if we can build a model that can successfully predict Ivies+ acceptances 75% of the time.</p>
<p>A few themes recur with high frequency. Sparkling essays, achievement in context, and unique combination of EC’s with superior accomplishments. These are hard to capture quantitatively, but I don’t think it is impossible. The solution may involve some compromise to how inputs are fed into the model. Need help here.</p>
<p>Perhaps you did not understand the term in context. I just meant what is not specified as essential on the admissions page. (In contrast to a h.s. transcript & a standardized test score or scores.)</p>
<p>Generally, achievements which do not come from within have difficulty being sustained. My children never did anything in the off-campus or e.c. category which I suggested first. And one of those children is still engaged in both of her two main e.c.'s after college. She did it overseas as well, of her own volition. It also becomes difficult to feign passion for something (particularly in an essay) that one is lukewarm about. Any peers of my children whose parents did push them relentlessly in those e.c.‘s (esp. when it was the parents’ idea to begin with) were ones who left the e.c.'s, and usually abruptly.</p>
<p>I suppose there are some Ivy admits whose off-campus accomplishments were largely or partly a result of parental “pushing,” but that is not always a bad thing if it’s a matter of encouragement, as well as rational persuasion to continue something (“go the distance”), such as toward a logical competition, toward the end of the age of qualification, etc. Plenty of students say later that they’re glad their parents “pushed,” encouraged, or persuaded them to stick something out.</p>
<p>
</p>
<p>But the slight difference is not enough to generate another “level” of acceptability, particularly when stacked against other (more important to Ivies) elements. Your earlier model had an enormous number of discrete levels of scores with points assigned to each level. There are ad reps on CC who have posted that this is not the way it works. Rather, again, it’s broad ranges, it’s context, it’s patterns that are looked at for any particular student, and it’s one element in a combination of desirable factors. There is not in this country an arbitrary weight assigned to an SAT score, at least for the Ivies.</p>
<p>PCP, I agree that it’s an unusual situation, but a kid who truly understands the research she/he is doing (or other passion) and who can teach/share it with others is not an opportunity limited to science fairs. Classroom teachers or those who sponsor ECs can see these interests and comment on it in rec letters. Choose those letter writers wisely!</p>
<p>And heck, he was still hosed by the Ivies. ;)</p>
<p>You can’t get admitted where you don’t apply, eh?
I’m not “upset” at all; I’m quite calm and just debating here.
The only point I was making was that even if you took all the top / perfect or close to perfect scorers, they don’t fill the Ivies because a) there are still more seats left over and b) many of those kids aren’t bothering to look at Ivies in the first place. I think some people are blissfully unaware of huge swathes of the population whereby a smart kid going to a good state flagship is a perfectly fine outcome, not a suboptimal one. I have no interest in sending my kids to our state flagship for many reasons, but I had to overcome my East Coast bias about state schools and recognize that many, many smart kids from many places go to state flagships and don’t even consider the Ivies. The Ivy madness is really concentrated in some parts of the country, and does not pick up all smart kids. That is all! .</p>
<p>The whole point is that you can’t, any more than you can look at an attractive group of women and predict which one you’ll fall in love with … any more than you can look at a set of colleges that meet certain criteria (size, urbanicity, academics, etc.) and predict which one your kid will fall in love with.</p>
<p>And attempts to try that will flop the moment they become widespread, PCP. That’s the whole POINT. Ten years ago, it was “sit-up-and-take-notice” when a kid built houses in Ecuador over the summer. At that time, that might have entered in your model positively as a wow factor. AS MORE KIDS DID IT, it became less wow and now these days might even enter as a negative (“rich parents bought kid an experience”). Don’t you see? It’s like trying to predict the popularity of a trend. The moment it’s overexposed, it loses its value. This is what you aren’t hearing.</p>
<p>Referring to the concept of weighting the 2400 from rich kid down and the 2400 from poor kid up, PCP said:
</p>
<p>I’m confused. Can you help me out here?</p>
<p>You built a model that explicitly gave someone x points for 2300, x+10 points for 2350, x+20 points for 2400. You did so because you believed that adcoms looked at those numbers out of context – that Poor Kid got his x points for 2300 and Rich Kid got his x+20 points for 2400 and the fact that Rich Kid was tutored within an inch of his life – well, that’s how the cookie crumbles. You were explicitly modeling something OUT of context.</p>
<p>And now you say that they don’t take context into account enough? MITChris was certainly indicating that they DO take context into account, which is why Poor Kid’s 2300 may indeed be better than Rich Kid’s 2400. </p>
<p>But what you’re not getting is that these adjustments are STILL judgmental and “feel” in the context of the entire app. It’s not as linear as adding 10 points and subtracting 10 … any more than your wife’s smile adds 10 points to her versus the woman next to her.</p>
<p>That’s usually a fallacious question, and in this case the situation is a bit sharper.</p>
<p>When:
– A (alone) is more influential than B (alone), and
– the relationship is one-directional (having strong A increases the chances of having strong B much more than the reverse), and
– A is the only commonly available measuring scale with this feature, and
– direct measurement or definition of an underlying True Causal Thing is impossible</p>
<p>then avoiding causal language is harder to sustain.</p>
<p>
</p>
<p>Where non-mathy measures are available they play exactly the same role.<br>
Perfect verbal score on the old SAT (or MAT or GRE), attaining fluency in several languages, placing near the top of the class at one of the top ten high schools or universities in the country, having memorized the Bible, monster IQ scores, etc all would be functionally equivalent to unusually high current SATs. Any such measure is informative when presented, but very few measuring scales apply to tens of thousands of people at once. For the college admissions there are test scores and a few of the science competitions, maybe chess ratings, and not many other possibilities.</p>