<p>
</p>
<p>Where’d you get that number from? If we assume that those who do not take the SAT are incapable of getting 2400 (though I’m sure there are a few dozen others who could), 2400 is above the 99.99 percentile among graduating high school seniors; that most nearly corresponds to an IQ of 156.</p>
<p>
</p>
<p>Yes, but as the percentiles indicate, “fairly easy” is probably not the most accurate diction.</p>
<p>An admissions Dean at a top school recently told me that they look at the SAT as confirmatory evidence that what they are seeing in the rest of the application is sound. Many of the high GPA great EC, etc. kids also have high test scores. If there is a big disparity, basically under 2100, then that is a yellow flag. It’s not that they are looking for kids with 2300s, it’s that many of the kids they are looking for also have 2300s. And, there is a difference.</p>
<p>
</p>
<p>That’s a misunderstanding of the correlation - it doesn’t work because you have no way of teasing out the other factors from the numbers. You can’t draw a conclusion as to probability based on a single factor without having a way to exclude other variables – and the colleges don’t make enough data available for you to sort out the other variables. </p>
<p>In other words, we don’t know the GPA of the kids in each ranges, but GPA probably correlates fairly highly to test scores. That is, on the whole, kids with 2300 scores probably have higher GPAs than kids with 2100 scores – but looking at stats for admissions for 2300+ won’t give you the information you would need to compare admission likelihood of a kid with a 2350 and a 3.8 uw GPA to a kid with a 2150 score and a 4.0 GPA. </p>
<p>Also – precisely because the colleges report super scores and retests are encouraged, the numbers you see tend to be inflated. Kids are encouraged to retake the test even if their scores were already good enough to get in, and colleges report only the best combination of scores. That system is designed to push reported scores upward - but that doesn’t mean that the the bottom line has changed. If colleges were required to report complete score ranges – the lowest as well as highest scores of all admitted students – you might see a very different picture. There is NO difference between the intellectual ability of a kid who takes the test one time as high school junior and scores 1980, and the kid who takes the test 3 times, initially scoring 1920 and then raising their score to 2250 after repeated sittings and test prep. In fact, if anything, you could argue that the kid with the one-time 1980 is smarter, since the first “raw” score was better. (Keep in mind that lower income kids or kids from public high schools are far less likely to retake the test, both for economic reasons and because they are in an environment that does not tend to encourage retaking when the initial score is perceived as fairly high). </p>
<p>For purposes of REPORTING it is in the college’s interest to report the highest scores, and to encourage students to think they need the best score possible. But for the purposes of ADMISSIONS decisions – if the goal was to bring in the smartest students – it really would be in the college’s interest to look at the lowest scores as well. Every college says that they don’t do that – that they look at highest scores only – but IF they value the score for predictive value, that makes little sense. Obviously they would be getting more accurate information if they looked at the history and pattern of test-taking.</p>
<p>
</p>
<p>Actually, you can. noimagination communicated the implication of the statistic completely correctly. Based on what the school has told us, a student who is randomly selected who has an SAT score of 2300+ has a chance 2.23 times better than a randomly selected student with an SAT score between 2100 and 2290.</p>
<p>
There’s still a correlation. It just may not mean anything. For whatever reason, kids with higher SAT scores are admitted at higher rates. </p>
<p>Obviously, the chance is not random in either case. That may have made my wording a bit misleading.
You provide one compelling reason why colleges may in fact care a good deal about the difference between 2300 and 2400 - reporting those scores for rankings and/or propaganda.</p>
<p>
</p>
<p>Obviously, it’s a combination of causation and correlation. Nonetheless, if it were merely correlation, then doesn’t that indicate that it ought to be causal?</p>
<p>
Since it is in the best interest of a college to be able to report higher scores, doesn’t that inherently imply a desire to admit higher scoring candidates?</p>
<p>Edit: I see noimagination got to it before me.</p>
<p>^ It seems that both you and noimagination picked up on that. :)</p>
<p>
The SAT is not valid as an IQ test in any way shape or form, because no IQ test could be valid if it allowed retaking or preparation for the test. Whatever correspondence it may once have had no longer exists. </p>
<p>In other words… a person who scores 130 on an IQ test will probably score higher if they have the opportunity to review their answers on the test, practice with a bunch of similar tests, repeatedly practice the type of questions that give them the most difficulty… and then retake the test again. That person’s 3rd “IQ” test in 6 months might show a score of 150 – but that wouldn’t mean that the IQ had increased.</p>
<p>
No, I disagree on that point. Correlation never implies causation unless you can be confident that other variables have been neutralized.</p>
<p>
</p>
<p>You misunderstood my point; I said “ought to be,” not “indicates that it is.”</p>
<p>
</p>
<p>Yes, I agree. When did I argue that the SAT was a valid IQ test or that IQ scores do not change? I was merely attempting (though tellingly failed) to rationalize your seemingly arbitrary figure of 135. You brought up the IQ example.</p>
<p>
Ah, I see.</p>
<p>^^ the IQ example was used only as an example of threshold…the premise being that there is a threshold SAT score and after that threshold is met other factors come into play. The example was of an IQ being used as a threshold for a gifted program…once the threshold is met it is not as necessary to continue to parse out the higher scores.</p>
<p>^ Yes, it appears as though I read into the example too much. I suppose the number choice was intentionally arbitrary (and not as analogous as it could have been and, as a result, somewhat misleading).</p>
<p>Wow! I left on a business trip, now I’m in an entirely different city (OKC) and still find people arguing this topic, one which I’m sure has come up many times on this website. </p>
<p>I guess it all depends on what you mean by the term “blunt”. The most interesting book, statistics-wise I ever read on this SAT subject was “the Early Admissions Game”, and it would tend to lend some credence to silverturtle’s point about the perfect score havng some particular meaning, but it depends on how you interpret it.</p>
<p>To summarize for those who haven’t looked at it - In the book he collects a lot of data, primarily SAT data, from various high achieving applicants in an attempt to prove his thesis that there is a significant benefit to applying early. He produces one table where he shows, based on his data, what your chances are to get into a school with a certain average SAT, based solely on what your SAT score is. If I recall correctly, for students with a perfect 1600 (it was the old test), they stood around a 50% chance of getting into a school with an average SAT of 1500-1600 (aka HYPMS). It dropped off pretty precipitously from there. I obviously don’t carry that book around with me, but I think it dropped to around 28% for kids with 1500 or above SATs. Now, you can read it two ways - for a school with an average acceptance around 10%, increasing it to 50% merely by having a high test score (all other things equal) is pretty remarkable statistically. OTOH, since we know there are very few perfect scorers, it is also pretty remarkable that it is still only basically a coin toss at that level. Below that perfect level, even at very high scores, SAT was a helpful piece of informaiton, but not so much so. Anyway, it’s sort of an interesting book, although he conflates mean and median (and admits it), which is a little annoying. </p>
<p>One thing I am glad got dropped pretty quickly here is this discussion of the Flynn effect. I don’t think it is germaine to this SAT discussion at all. As far as I know, the Flynn effect looks at increases in the average IQ, and many, if not most researchers believe those increases are primarily due to major increases one or two sigma below the mean - not at the top levels we are talking about here.</p>
<p>I also was interested in the discussion of the old SATs (actually, there are several versions even I am aware of) versus the new SATs. I don’t really know which tests were easier or harder, but I do know they are different. The new math test actually has one thing I really like a lot - the “fill in the blank” problems. But it got rid of the quant comparisons, which I always thought were sort of tricky. Since people can use calculators now I bet they felt they had to eliminate problems that could come down to who could plug numbers in faster.</p>
<p>The verbal test is so different they even changed the name- to critical reading. Is it easier or harder Again, I don’t know. But I always thought analogies and antonyms were tricky. </p>
<p>Obviously, I can’t compare the difficulty now, I’m an adult with a few college degrees under my belt. I didn’t get a perfect score as a kid, and I am sure if I were a kid I wouldn’t get a perfect score now.</p>
<p>Sorry for the long post.</p>
<p>^^^
Oh, and to add to an already overly long post.</p>
<p>For those SAT tables, there was also a threshhold on the lower side, as many people have pointed out, where it was almost impossible for anyone to be admitted with a score like that.</p>
<p>So it was like a function with two major discontinuities - one somewhere around 1400, and one at 1600. In between, it didn’t seem to matter much.</p>
<p>Just another brief note on mifune’s theory. He references this :
<a href=“http://www.infogoaround.org/CollegesChinese/RevealRanking.pdf[/url]”>http://www.infogoaround.org/CollegesChinese/RevealRanking.pdf</a>
which says only this about SATs
"Examine MIT admissions in Figure 1. The probability of a student’s being admitted
rises steeply and monotonically in his or her combined SAT score, suggesting that MIT is not engaging in strategic admissions. Now examine Harvard admissions in Figure 1. The line has a flat region that suggests that the probability of a student’s being admitted is about 10 percent regardless of where his SAT scores in the range between the 93rd and the 98th percentiles. Above the 98th percentile, a student’s probability of admissions rises steeply. Finally, consider Princeton admissions in Figure 1. At Princeton, the admissions probability rises to 20 percent at the 93 percentile, then falls to 10 percent at the 98 percentile (precisely the region where competition is toughest), and then rises again for students with SAT scores in the top 2 percentiles.
In short, it appears that Princeton practices more strategic admissions than MIT or
Harvard. When we see the revealed preference ranking later in the paper, we will see that
Figure 1 makes sense because Harvard and MIT could benefit less from strategic admissions
than Princeton could. While Figure 1 is not definitive, it provides suggestive evidence that
even a highly prestigious school may practice potentially costly strategic admissions. "</p>
<p>Now, this is the only reference to non-median SATs. And the archived CB data for 2004 shows that the top 2% of scorers in both math and verbal scored 750-800. I suggest that this study simply is using that metric and not making any attempt to discriminate between scores at or above 750. This seems reasonable to me because there is no other mention of non-median scores, and because such discrimination would not serve the purpose of the paper. Therefore, it would appear that any attempt to extrapolate about scores within the 750-800 section using this study is on exceedingly shaky ground at best.</p>
<p>^ See post #134.</p>
<p>silverturtle - I just wanted to make it clear to those who have not read both mifune’s thread and the cited study what was there and what wasn’t. I understand you are letting go of mifune’s assertions.</p>