<p>:-o!!</p>
<p>TAKE THAT BACK!</p>
<p>Agree with the argument that there is little difference between 2200 and 2400. </p>
<p>My original post asked: </p>
<p>Is it true that the 2400s only know the tests and nothing going on for them when compared to these 2200s?</p>
<p><em>giggle</em></p>
<p>:)</p>
<p>Hmm, I think that really confirms me in not applying to Caltech. Really, Ben, this seems a bit ridiculous. Do you <em>seriously</em> consider a 800 better than 750 on the SAT math section? Are you aware of the fact that 750 converts to only two false answers?
[quote]
To all the pious defenders of MIT's declaration that 2250=2400, I make fun of you.
[/quote]
I make fun of all those that think that scoring 800 is more impressive than 750. Oh my ...</p>
<p>aw5k -- you missed the point. The point was that the ways of adjusting for the meaninglessness of the 50 point gap are often much worse and introduce more noise than if we took it at face value.</p>
<p>"Is it true that the 2400s only know the tests and nothing going on for them when compared to these 2200s?"</p>
<p>look. for the most part the really really smart kids will get at least pretty damn close to a 2400. but they will, by virtue of their intelligence, not rely on their SAT scores in the least to show it. Here, two examples: student A is smart but not that smart, studies really hard and gets a 2350 on their SATs, suddenly, they're above the average SAT range for Yale and they figure, WOW, I can get into Yale! So they apply. Student B has always been intellectually curious and very very intelligent. She naturally then has on her resume a slew of cool extra curriculars and great recommendations and college courses and projects and awards and XYZQ. She takes the SATs as a formality, studies a bit, gets a 2200, and decides to apply to Yale anyways. I'm going to tell you what happens in college admissions heaven. Student B wins (reality is a bit more foggy). Test scores are a flaky thing. They can't ALTER your application in any real way, they can only support it. It's really depressing to see people weighing them so heavily. What are the chances you think we'll really miss a brilliant kid by rejecting a 2400 with little to no corroborating evidence in the rest of his application?</p>
<p>I just think that getting a super-high SAT score often correlates more strongly with preparation than with actual native ability. I studied my hind end off for the GRE, and ended up getting a perfect score in math, while I prepped nada for the SAT and got a 690. I find it difficult to believe that four years at MIT made me better at high school math (if anything, I think my addition skills have gone down the tubes!).</p>
<p>
[quote=Ben G., who is baiting me]
Sure, SAT's measure imperfectly, but (surprise!) so do all your other touchy-feely measures of ability and potential -- and the latter are more prone to manipulation (on the admissions end) for reasons having nothing to do with merit or promise.
[/quote]
I've actually come to the somewhat hopeless and nihilistic opinion of late that it's actually not possible to pick the "smartest"/"best" 1500 students out of an applicant pool that's clearly quite qualified. Success in science and engineering has to do with a combination of your native abilities and your sheer force of will, and I've seen plenty of perfectly smart kids stumble at MIT because they didn't have enough on the hard work side. I've also seen plenty of not-geniuses really excel because they had an extraordinary ability to dig their heels in and work work work. So regardless of what high test scores measure, I don't think MIT should only accept people who have lots of whatever it is, because I haven't seen it correlate well with success at MIT.</p>
<p>For that matter, because I am a biologist, and a statistics lover at that, I can't help but see error bars in SAT scores. If everybody took the SAT an infinite number of times with no preparation allowed, I would be more happy (though not totally happy) with saying the SAT measured something with some degree of accuracy. But as I see it now, the error bars on individual scores are just too big to say that someone with a 2400 is "smarter" than someone with a 2250. The overwhelming likelihood is that the theoretical smart of the two people are equal, or at least equal to the degree that we can measure them.</p>
<p>For that matter, I find it troubling that standardized test scores correlate suspiciously well with purely social attributes. There was a link in the Parents Forum the other day ([url=<a href="http://www.udel.edu/educ/gottfredson/reprints/1998generalintelligencefactor.pdf%5Dhere%5B/url">http://www.udel.edu/educ/gottfredson/reprints/1998generalintelligencefactor.pdf]here[/url</a>], on page 5 of the PDF) showing correlations of various IQ ranges with the divorce rate, poverty rate, childbirth out of wedlock rate, and the rate of welfare receipt. This bothers me quite a bit -- there are a lot of social issues tangled up in standardized test scores, and you can't just swallow a score without thinking about the background issues which influence getting a good score on it.</p>
<p>Let me make myself clearer since I see some arguments against my earlier notes based on misinterpretation/misunderstanding:</p>
<ol>
<li> I am talking about the test scores in general, not in particular cases. So those who argue a 2200er could very well be equivalent with or even smarter than a 2400er is totally off the track. Of course you can find plenty of such cases, especially if you compare a Non-English speaker with a native speaker. Whats your point? My question to these people is: can you deny the fact that in average, 2400ers are in higher scholastic aptitude level than 2200ers? If not, why bother with SAT at all? According to your logic: If 2400=2200; then 2200=2000; then 2000=1800; then 1800=1500
eventually we will get 2400=0!</li>
<li> Even though I am not a perfect scorer and dont even dream to be, I feel very offended by someone comparing a perfect SAT scorer to a lottery winner. That only confirms one thing: there are all kinds of people in this world, CC is no exception. Fortunately most people would just ignore such crap comments which is based of ignorance, jealousy, prejudice or whatever it is.</li>
<li> I like to remind people one thing: 2400 is the cap of the test score. So the most important difference between those 2400ers and 2200ers is this: the former could reach a much higher level than the later! </li>
<li> Again, test score is not, should not and never has been the only indicator of applicants success in college admission or in their future careers. But thats not saying a perfect score doesnt matter. Statistics shows perfect scorers have much higher acceptance rate than any other scorers in selective colleges, including MIT. Why? Not because the score difference , but because over all, the perfect scorers tend to be exceptional students, period. So those who quoted Ben Jones to argue with me really dont get the point at all. He wouldnt be Ben Jones if he makes admission decision on "hair-splitting" SAT scores. You know what, nobody does! If you really want Ben Jones to give you an authoritative answer to this issue, ask him this way: everything else equal, would you pick a 2400er or a 2200er?</li>
</ol>
<p>I can guarantee that benjones would say there is no way "everything else" could be equal.</p>
<p>hey, 2400~2390, right? By your same argument, 2400 being close to 2390 implies 2400 = 0? No. First of all, the distribution of the SAT is about normal implying that there is a small difference between 2200 and 2400 (maybe, like,2-3%) and a larger difference between 2000 and 2200 (maybe 6%). It goes up from there. </p>
<p>Secondly, the SAT measures basic skills, so getting above 2200 should be enough to demonstrate that you're not a complete imbicile. There are just better ways to determine potential.</p>
<p>To those who dont know the difference between gaining perfect SAT score and winning a lottery:</p>
<p>A dog would have chance to win the lottery; but a dog, and its kind, would never never never never never have any chance to gain a perfect SAT score. Get it?</p>
<p>ivyapp -- I hope you're convinced that you and I are fellow travelers in at least some ways, so my criticism here is not intended to hit you too hard. But here is what you are misunderstanding about the lottery analogy.</p>
<p>Let's agree that there's an underlying raw ability r, which is an unobservable attribute of high school seniors (let's agree to measure it in SAT points, just for numerical convenience). It's unobservable because we can't measure it directly and precisely, but we can measure it, as long as we are willing to accept some error. In particular, what you get is a particular SAT score, </p>
<p>s = r + e</p>
<p>where e is an error term which captures how much this SAT score is off from the true raw ability. If you get unlucky and your mind is not working so well that day, e will probably be negative, so the SAT score will underestimate your true ability. If for some reason your examiner gives you a few extra minutes by accident, e may be positive. But the point is that e is random -- it depends on accidents of fortune unrelated to pure ability.</p>
<p>Winning the lottery means having your e be high, so your SAT score is equal to or greater than your raw ability. There's nothing special about that -- you just got lucky in having the measurement not screw you over. On the other hand, if e is negative, your SAT score is low compared to your true raw ability.</p>
<p>On average, repeated administrations show that e is about 60 points in magnitude, so (roughly speaking) you can hope to estimate ability within an 120 point range, but not more precisely than that.</p>
<p>Small example: suppose real raw ability is r = 2340 is the same for two people. One gets lucky and gets a lottery draw of e = 60, so his score as measured is 2400. Another gets an unlucky lottery draw of e=-60, so his score as measured is 2280. But they are actually the same. They differ only in the meaningless and random error terms.</p>
<p>So of course r is serious and real and important, but e is a lottery, and that's why people say that getting a 2400 vs. 2280 is a matter of a coin flip and is not informative.</p>
<p>To mootmom:</p>
<p>I trust Ben understand what is a hypothetical question. But for you, I am happy to restate the itlike this: everything else similar, meaning no significant advantages one over the other, would you pick a 2400er or a 2200er?</p>
<p>Now what I would say to Mollie, who clearly knows a lot of statistics, is that a random draw from a distribution with an unknown mean is optimal estimate of the mean (in the Bayesian sense, or in the minimizing mean-squared error sense, whichever you like). That means in some sense that you can't do better than taking the draw you get to be your best estimate of raw ability.</p>
<p>So if you see a 2400 and a 2280, you should assume the 2400 has more raw ability, and you'll be right most of time time (how much of the time depends on the precise distribution of the error term). Want to argue with that? (I think it would be delightful if we turned this thread into a statistics debate.)</p>
<p>While we're at it we should probably require the scoring sheets to see if the applicants 7xx came from missing an easy, medium or hard question. Is there a difference between an 800 and a 760 if a begining of section no brainer question was missed? How about if it were one of the hard complex end of section ones? It is easy to correlate SAT scores and student ability if you know them well (your DS or DD for example), in a college application the other factors take on more importance.</p>
<p><a href="I%20think%20it%20would%20be%20delightful%20if%20we%20turned%20this%20thread%20into%20a%20statistics%20debate.">quote</a>
[/quote]
Yay, me too! :D</p>
<p>I guess I'm thinking of the score-potential relationship in that I'm assuming that each score has an approximately normal distribution of potential-that-you-can't-really-measure. My feeling is that the distributions overlap substantially for 2400 scorers and 2250 scorers, but not so much for 2400 scorers and, say, 1800 scorers.</p>
<p>And to me, that means the SAT is a handy tool for telling the difference in potential between someone who got a 1500 and someone who got a 2400, but I have a lot less confidence in its ability to distinguish between the abilities of people in the top ~1% of the college-applying population.</p>
<p>
[quote]
So if you see a 2400 and a 2280, you should assume the 2400 has more raw ability, and you'll be right most of time time (how much of the time depends on the precise distribution of the error term). Want to argue with that?
[/quote]
No, I don't -- I think that's what the higher 2400-vs.-other-scores admisison rates are telling us. When you select for a variety of imperfect measures of potential, the 2400 scorers are selected at higher rates than non-2400 scorers. I just object to the idea that therefore all 2400 scorers should be admitted without respect to other apsects of the application/intellectual life of the applicant -- that the test score tells you something you can't figure out in other ways.</p>
<p>As a sort of aside, I was thinking about graduate school admissions, and how different they are from undergrad. A graduate school admissions committee, at least in biology, would never dream of letting a 1600 (perfect) GRE scorer with no research experience into a top program, but would happily give a spot to someone with 60th percentile scores and outstanding experience. Nobody bats an eye at this. I'm not sure why test scores are considered to reflect some deep truth about an applicant at the undergraduate level rather than the graduate level -- maybe because more people at the undergraduate level get perfect or near-perfect scores?</p>
<p>As a further aside, I am quite curious to hear exactly how a dog could win the lottery. A duck could tell the cashier to put the ticket on his bill, but a dog? I'm not sure.</p>
<p>I know that people have commented that a lot of the kids who get in seem to be doing research or something else. I don't think a lot of people get that research is a much much better indicator of how you will do in college than SAT scores, enough that good research makes SAT scores worth nothing (assuming you have like 2000ish+). Research takes months of persistance, working on cutting edge work, intense thinking, etc, all of which are great indicators of college success.</p>
<p>I like Ben Golub's s = r + e equation. That's pretty accurate, I'd say.
That said, I think that the SAT CR and W section really are just about learning by heart. I am 100% sure on this, from my own experience. I am an international applicant, the first time I took the SAT I got a 640 on CR. That seems not too bad for an international, but not impressive at all. Well, next time I studied all those SAT words (how boring and dull!), and I got a 770. Wow, that's great!
Well, I am not smarter now than I was when I took the SAT for the first time. If we stick to the s = r + e equation, then it's not just the 'e' that changed, I didn't have just more luck - I just knew more words. So, I am really very positive that the SAT CR and W sections might measure your potential t(to think critically, blabla) to a certain degree, but this potential is totally blurred by one's need to study those SAT words.</p>
<p>aw5k, that's the very reason I am suspicious of taking the SAT as absolute; there are students out there (take myself from a year ago) who do not <em>know</em> that there are resources available which can be used to substantially increase SAT scores. These resources (books, etc.) can have a large impact on the "e" term - either increase or decrease, depending on your take on whether only preparation-less testing measures "true potential." In either case, however, those students who do know about such resources are given the advantage since the college application does not ask, "How many hours did you prepare with the ACME Study Guide before taking the SAT?"</p>
<p>
[quote]
As a further aside, I am quite curious to hear exactly how a dog could win the lottery. A duck could tell the cashier to put the ticket on his bill, but a dog? I'm not sure.
[/quote]
</p>
<br>
<p><em>< ></em>< >_<</p>
<br>