<p>The real problem with the CDS is that it shows the profile of (either admitted or matriculating, I forget which) but it’s not showing the profile of the rejected. Using ACT scores just to make it clearer, it would be one thing if you said that College A’s class 25/75 percentile was 30 - 33 and the rejected were 25 - 28. But for all you know when College A reports that the admitted / matriculating class has 30 - 33, maybe those rejected have the exact same profile. You don’t know.</p>
<p>completelykate for the win! Now, see, if I were an adcom – that’s <em>exactly</em> the type of spirit and sense of humor that would make me stand up and take notice.</p>
<p>I find this continued emphasis on it’s all about the scores and not about the personality completely at odds with what the colleges actually do, and for the life of me I don’t understand why what they say they do isn’t taken at face value.</p>
<p>
</p>
<p>There isn’t a university in the United States that admits fewer than 40-60 percent of perfect SAT scorers. Several of the top 12 schools publish admission data on their web sites and in their student newspapers. Have you looked? For example,</p>
<p>[Brown</a> Admission: Facts & Figures](<a href=“Undergraduate Admission | Brown University”>Undergraduate Admission | Brown University)</p>
<p>Getting 800 on any given section of the SAT puts you in a 20-24 percent admission category at Brown. Perfect score (36) on the ACT, which is more common than 2400 SAT, is associated with 30 percent admission rate. Valedictorians got in at 25 percent. </p>
<p>If you know enough math you can make estimates of the rates for people with any given combination of the SAT scores and class rank categories for which the data are provided on the web site. It’s a bit involved, not an off the shelf statistical problem by any means, and people with the necessary skills generally have better things to do, but I assure you that in principle it is very much possible to derive information about the multi-factor admission rates from the public information. </p>
<p>Single digit admission rates are a fiction. The reality is that half or more of applicants to the uppermost schools have zero chance of admission, and that can be predicted in advance. Those with credentials suggesting that admission is not a ridiculous idea, have chances in the 10-20 percent range and this can be adjusted up or down in individual cases as modeling skill permits.</p>
<p>
</p>
<p>I didn’t see OP’s model proposal, but yes, complex models (given the data resolution available) are not likely to be good. However, a LOT of information can be had, both quantitative and qualitative and a good idea of how they might interact, from published data. It is not an efficient use of time for people who usually model more interesting or lucrative things, but there’s no reason the same technology can’t be applied to college admission. Admission is actually better suited to quantitative modeling than many other things for which models are developed.</p>
<p>
</p>
<p>I’m not sure I agree but let’s assume you’re right … so I figure out my kid’s chance is not the basic 10% but is actually 35% (or 5% for that matter) … so what? Given the results what different strategies would I recommend for my child? The only actionable benefit I can imagine is possibly adjusting the # of apps sumitted … but frankly I would have no where the confidence in the model to lower my lid’s number of apps because a model predicted s/he is in good shape (I start with a bias to many apps to low acceptance rate schools). Bottom line I’d rather my kid submit the extra apps and I’d rather spend my time worling an extra few hours to pay the adiditional application fees than play around with a predictive model.</p>
<p>Many of us have argued that “Ivy caliber” covers a wide range of students, most of whom will not be admitted to the very small number of schools which PCP and POIH deem to be Ivies+ (the discussion seems to relapse back to HYPSMC with dismaying regularity) if only because of the limited number of spaces (in particular dorm space) available. But both have argued that the proof that the formula is valid is admission into these Ivies+.
Except for Caltech, which seems to rely on pure stats more than other schools, it is hard to predict outcomes for individual students. An example may be PCP’s own kid, who got admitted into Caltech and Waitlisted at MIT. To me, being WLed at MIT is proof that the applicant is Ivy+ caliber (never mind the Caltech acceptance as well). But PCP and POIH would like the formula to be more predictive. Will we ever know why PCP’s kid got WLed at MIT? It cannot be the stats, which were good enough for Caltech! So it probably is some other factor that is less easily quantifiable. And it may not have anything to do with the student’s profile at all, but something totally outside the student’s control, like a member of the adcom willing to go to bat for the student.</p>
<p>
</p>
<p>Exactly. Unless the chances can be predicted to a high level, there is no point in doing the exercise. It does me no good to know that my kid’s specific chances at a selective school are 10%, 15%, 20% or 30%. It may make ME feel better, but … it’s still only 1 in 3! Until it’s close to a sure thing, I don’t see the point.</p>
<p>
</p>
<p>I know my way around statistics – CHAID, discrim, regression and all the rest. It’s part of what I do for a living :-). However, you just can’t model things that have a large component of that qualitative, it-either-grabs-me-or-it-doesn’t component.</p>
<p>To MITChris’ comments upthread, one <em>could</em> actually weight, say, a SAT score by the economic achievement of the parents, and weight down, say, the 2400 gotten by the upper middle class and weight up the 2400 gotten by the poor kid. That’s all feasible if you used some kind of proxy for parental income - maybe education or zip code. However, what you aren’t hearing is that a lot of it is indeed a subjective, does-this-kid-strike-me-as-interesting-or-not. That simply can’t be modeled, any more than you can model who you’re going to fall in love with. We all know people who said they preferred blondes who wound up with redheads, people who said they preferred tall people who wound up with short ones, etc. Because there is a spark, a something extra, a something that makes someone sit up and take notice. And here’s the thing. The thing that makes Adcom Member A sit up and take notice may or may not be the thing that makes Adcom Member B sit up and take notice. </p>
<p>Look what happened with the whole build-houses-in-Ecuador thing. Parents derived a model in their own heads that their kids had greater chances of acceptance if they showed that they did volunteer work in poorer countries. So they all sent their kids packing to wherever. And the model broke down because as soon as the adcoms realized that they were “rewarding” that one behavior – that is, “living by the model” – they decided they didn’t want to do it anymore.</p>
<p>Look – for those of us who are going or have gone through the “visit colleges and see what your kids think of them” stage …
Do you think you could have built a “model” of what colleges your kids would like?</p>
<p>Sure, you might be able to put in that generally speaking, kid has a preference for small vs large, rural vs urban, no Greek life vs Greek life, big sports scene, etc. And maybe some things come off the list entirely (no NYU for the kid who wants a bucolic rural scene).</p>
<p>But is there anyone here who hasn’t found that a college they thought their kid might like “by the numbers” – the kid didn’t like? Or vice versa?</p>
<p>And why do you think that is? Because when you visit a campus, you GET A FEEL. Could these people be my tribe? Do I see myself in this environment? And you react to that FEEL, not just the numbers or attributes.</p>
<p>Don’t you get that just as the kid looks around and says “Do I see myself here?”, the adcom looks at kid and says “Is this a kid I want to know / have come here?”</p>
<p>
</p>
<p>PCP’s son was ALREADY IVY CALIBER the moment he got accepted into Chicago EA. He didn’t need to wait for Caltech to prove that, and whatever happened at MIT (whether WL or denied) doesn’t change that.</p>
<p>
</p>
<p>It is not a standard statistics problem amenable to techniques at that level. Which is not to say that it can’t be done or that running the standard toolbox wouldn’t be informative.</p>
<p>
</p>
<p>The “large” component of randomness may turn out to be fairly small. Reconciling the high application volume, low admission rates and substantial overlap in applicant pools does require that some randomness be present — but one can (in principle) derive quantitative estimates of how much, and it’s not necessarily all that big. </p>
<p>
</p>
<p>Interesting or not can be modeled in all sorts of ways, especially when considering a single applicant rather than trying to build a comprehensive analysis of all applicant types. </p>
<p>Also, you can quantify whom you are likelier to fall in love with. Sometimes there really are big and unexpected discrepancies between a husband and wife’s height, education levels, languages spoken, or religion. But they are uncommon. </p>
<p>Opinions differ on the value of admission probability models for particular applicants, but such models have implications for understanding the process as a whole across multiple schools, and this is of greater interest (Early Admissions Game study, Espenshade’s work, and others).</p>
<p>
</p>
<p>There are schools that reject 60% of applicants with perfect SAT scores? You’re kidding. I never would have thought it was that high.</p>
<p>PG:</p>
<p>Very true. I had forgotten the Chicago EA admission!</p>
<p>(continuing from post #7510</p>
<p>Your point, of course, was that a lot of kids with perfect SAT scores are admitted to the top schools.</p>
<p>It is dangerous to look at a statistic like that and draw the conclusion that getting perfect SAT scores will mean you’ve got a 40 - 60% chance of being admitted to a top university because the correlation isn’t necessarily there. Kids who get perfect SAT scores tend to be high achievers in many areas, it wasn’t only their SAT scores that got them into the top schools, it was many things. So getting my kid to get perfect SAT scores won’t make any difference in his chance of being admitted to a top school, because the many other parts of his application will stay the same.</p>
<p>s-rune said: “There isn’t a university in the United States that admits fewer than 40-60 percent of perfect SAT scorers.”</p>
<p>Even if true, in 2008 that observation applied to precisely 294 individuals. There were about 13,000 frosh slots at the Ivies. There were over 100,00 applicants who are yearning for a magic model that will tell them that they are more likely to be admitted to HYPed schools.</p>
<p>Two points:
1 - Models are useful to the extent they are generalizable. If it’s predictive for only a small percentage of the intended audience it’s close to useless.
2- If a model predicts that “perfect SAT scores increase one’s chances of acceptance” that would be scientific validation of the obvious; we already know that.</p>
<p>Kei</p>
<p>It seems to me that there are a number of scales upon which applicants are evaluated, and being at the top of certain of the scales is probably more predictive of admission than others. Thus (for example) I think it’s obvious that getting a 2400 on the SAT is (considered alone) better than being a valedictorian. Winning a top science prize is obviously very helpful, pretty clearly better than having a really good alumni interview. But how helpful is it to have one of the best essays of the year? Or to have unusually good recommendations? It’s very difficult to say, and it may vary a lot from school to school, and even from admissions officer to admissions officer.</p>
<p>PCP, my kid was waitlisted at Caltech and acccepted EA at MIT two years ago. There is no accounting for what they are looking for!</p>
<p>I will say that after S1 presented his research project at MIT senior year, several profs came up to him and asked if S had ever considered teaching. They knew enough to know he had done an outstanding job of explaining some VERY complicated concepts to a lay audience. Since S would LOVE to teach, this was music to his ears, and frankly, he didn’t care whether or not he advanced after that. He felt validated. At that point, we felt like he’d crossed the hurdle to getting into MIT. So, yes, he had the hardware, the scores, yada yada – but the MIT folks also saw the passion and depth of understanding. S was very fortunate to have had that opportunity.</p>
<p>Before I share my musings of the day on this thread, I’d like to openly apologize to epiphany for not realizing she had sent data to me for more than a week. My oversight.</p>
<p>
</p>
<p>But it is never considered alone: that’s the entire point. Someone who’s Val of a very rigorous private whose quality is known to the Ivies, and wth whom he or she is virtually evenly matched in ability with about 8 other students in his or her class, and which “someone” is also highly accomplished outside of campus (again, better than close competitors) has been proven over the last 6-7 years to be more valuable to Ivies than a 2400 scorer with little else to brag about. There are legions of examples of this; and some of these examples have been published in books.</p>
<p>And the reason the Val of the fab private would be more valued with a 2300 (or even less) than a 2400 competitor? Because the bottom line is that it’s about 2 things, mainly: (1) the four years of academic history and the content of those academics (which a 2400, in itself, does not show); (2) the student’s independent drive in noncompulsory factors, which is not manifested by a 2400 score (a compulsory admission factor).</p>
<p>It’s about permutations and combinations, but with a clear preference on the part of Elite admissions committees.</p>
<p>Again, we’re talking Ivies/Ivy-level here, the subject of the thread. The 2400 scorer is, in many other college environments, a walk-in.</p>
<p>
</p>
<p>I don’t know how dangerous it is. First of all, the 40-60 percent is an average of the rates for all 2400 scorers, many of whom don’t have high grades or achievements. For those who are also high by other measures, the rate of admission has to be higher than this average. </p>
<p>Second and more importantly, extremely high SAT scores essentially are causal of the other achievements. Those scores can’t be generated at will or through coaching, and students with perfect grades usually don’t have perfect test scores, but the reverse is often true.</p>
<p>Are they causal, or correlated with?</p>
<p>This whole thread also seems somewhat tilted towards mathy-sciency kids where there are more quantifiable measurements of achievement (winning an Intel award, etc.).</p>