<p>Actually not, because colleges never report the cumulative scores. The report the separate scores, and it doesn’t matter for their purposes if any given student fills all 3 high end slots. </p>
<p>What does matter is that when the scores are broken down, that the student offer some scores in the high end range. “Range” is all that matters – for example, if Yale’s median CR range is 700-800, then a 780 CR is not better or worse for them than a 720. What is important to Yale simply that 75% of the CR scores overall are 700+.</p>
<p>Let’s look at 2 hypothetical students: </p>
<p>Student A: (2070)</p>
<p>690 CR
690 Math
690 Writing</p>
<p>Student B: (2030)</p>
<p>710 CR
610 Math
710 Writing</p>
<p>Student A has a better combined score, but Yale gets the most value added by admitting Student B, with 2 scores above 700. So student B is contributing 2 midrange scores, whereas Student A falls in the “bottom” 25% range on all 3 counts. So to the extent that Yale wants to ensure that it keeps its test score numbers up – they are going to admit Student B over A.</p>
<p>Similarly – if they have student C, with 3 scores of 710 (2130) – and a student D with a pattern of 790, 790, 690 (2270) – the student with the lower cumulative score is more valuable for statistical reporting purposes. </p>
<p>Whether Yale or any other college employs that sort of analysis – I don’t know. I’m just pointing out that to the extent that test scores influence rankings, it is more effective for them to do so – and to simply categorize students by their score ranges rather than individual scores. (That is, to code the files based on score ranges rather than specific scores).</p>
<p>The SAT was designed to have a mean of 500 with standard deviation 100. That means that the max score would be 3 standard deviations above the average. As it works out, the averages are a little over 500, and the standard deviations are a bit over 100. So with an IQ std dev of 15, it would be fair to say that the SAT tops out around 140.</p>
<p>If you add in the somewhat arbitrary nature of 750+ scores, you essentially get a test that tops out (in terms of accuracy) at 135.</p>
<p>Well put, ThisCouldBeHeavn. And, just as the commonly used IQ tests become of little value at the high end, so, in my view, do the SATs. Neither can discriminate well above a certain point.
An extreme example to make the point - consider those reading achievement tests you took in lower school (K-6). They were pretty good at helping the teacher sort kids into reading groups. But if you gave those same tests to a bunch of adults, they would do little except to tell you who was close to functionally illiterate. The average high graduate, the smart college student and the guy who just won the Nobel for literature.
(Now, of course, how SATs are used is another question.)</p>
<p>I’m not at all sure what you are getting at in posting 164, silverturtle. Are you returning to the line of linking IQ and SAT scores? I thought that had been dropped since references to IQ testing were initially only for (I believe) illustration of the innate limits of std dev based testing.
The link between IQ and modern SAT score is modest at best. There are quite a few areas of intelligence that the SAT does not touch on at all (which are part of standard IQ testing). Also , the fact that SAT scores can be improved by prep implies that it is not strongly related to IQ. Your IQ certainly does not improve because you have memorized some vocab, for example.</p>
Um, I’m pretty sure you reintroduced the topic of a comparison between the SAT and IQ tests before Silverturtle. That is if I am correct in assuming Post 163 occurs before Post 164.</p>
<p>posting 140 by calmom brings up the concept of the SAT as functionally equivalent to an IQ test that tops out at 135 - not, I believe, asserting that a top SAT is equal to an IQ of 135, but rather (and in my view, rightly) noting that an IQ test that tops out at 135 does not allow one to discern IQs above that score.
posting 141 by silverturtle asks for calmom’s numbers and asserts that an SAT score of 2400 corresponds to an IQ of 156 (I think this misses the mark entirely - I believe calmom is talking about how std dev based tests function, not how the SAT relates to IQs)
the line is dropped until calmom’s posting 149. There she rightly asserts that the SAT is not valid as an IQ test.
silverturtle responds in posting 152 , agreeing that SAT is not a valid IQ test, and asserting that he was attempting to rationalize her 135 figure.
momofthreeboys resonds in posting 154 , noting that IQ was used as an example to support the premise that there is a threshold SAT score.
silverturtle responds in posting 155 , noting that he appears to have read too much into the example.
ThisCouldBeHeavn chimes in at posting 162 with a commentary on mean and standard deviation , stating that using IQ std dev and mean numbers, would mean that the SAT tops around 135 or 140. It is not entirely clear whether this poster means that IQ is related to SAT score - I think it means that a top SAT score is the mathematical equivalent of an IQ score of 140.
I come in at posting 163, agreeing with the numerical analysis of posting 162, and giving another std dev, ceiling limited example.
silverturtle comes back, posting 164 asserting “A given section tops out at 140 perhaps. But the overall test does not.”
I then ask , in posting 165, if silverturtle is asserting that SAT is related to IQ or if he is talking about std dev math modeling ceiling limits and noting that SAT is the same as IQ.
I missed part of this thread and didn’t see silverturtle’s posting 152.</p>
<p>Ok, but given the absence of data wrt to the overall standard deviation, the limits to each of the two relevant sections gives an indication of the overall limits of the test.</p>
<p>If the two sections are uncorrelated, the limit approximates to 155; if there’s a fairly strong correlation, the limit approximates to the mid 140’s. In the first case, you can take off about 10 points in terms of accuracy for the vagaries of near-perfect scores. In the latter case, you could take off probably 5-7 points.</p>
<p>Which I think reinforces my point back at post 94. The SAT is even more of a blunt tool now that score choice is an option for some students. Taking the SAT multiple times negates any possibility that the score can be seen as approximating IQ. One and done with a very high score, before score choice for SAT’s, gave colleges a valuable piece of information. If the schools can’t use it as well to differentiate among the top students, a 2400 has to be less valuable in the upcoming admission cycles.</p>
<p>I was perhaps too general in my phrasing; I intended to communicate that the IQ “conversions” were irrelevant. However, the fact that relatively few students “top out” on the SAT is quite contradictory and relevant to the premise of the thread.</p>
<p>You’re taking “top out” far too literally. There are many people scoring above the accuracy limit of the test. That’s what constitutes topping out.</p>
<p>Assuming that exact scores over ~750 tend to be arbitrary (and I’m sure that if you disagreed you would have said earlier because it was pretty clear that that point was relied upon several times), then it is fair to say that overall scores past a corresponding point are also somewhat arbitrary. </p>
<p>Assuming that overall scores over a certain point tend to be arbitrary (and again, this point has been made several times), than the differences between scores in that range would be meaningless or at least not an effective predictive tool.</p>
<p>Assuming that the differences are not meaningful (a point that, again, has been made upthread), it doesn’t matter that relatively few people are scoring at any particular point in that range.</p>
<p>This fellow wasted a bunch of time researching old comparisons and developing his own criteria for comparison between IQ and standardized tests. I have no idea who he is. Frankly I think it’s a bunch of nonsense and wasted effort.</p>
<p>One thing he points out fairly consistently, is that Mensa no longer accepts the SAT test for membership (since 1994). This I know is true. Probably because of all the prep work and score choice. But I’m personally not a big Mensa fan either - silly people patting each other on the back for genetics they had nothing to do with.</p>
<p>I have repeatedly indicated that I do not think that there is a threshold beyond which score increases do not positively correlate with higher abilities.</p>
<p>Okay, even if we choose to disregard that study, there remains a great quantity of statistics that expose the same generalized level of admissions behavior. This subsequently leads to the same basic conclusion – that colleges do indeed discriminate among scores at the very high end of the spectrum and find higher performances attractable – rather than irrelevant – for purposes of admissions assessment.</p>
<p>Also, to anticipate the possible charge of the lack of inherent causality in the above findings, there is a discussion regarding that very topic in the first link embedded within this post. It is commonly believed that the SAT fails to be considered beyond a certain score boundary in college admission, but there is little evidence to actually support that. Mainly, it serves as an indiscriminate, misleading, and erroneous assumption rather than a substantiated argument.</p>
<p>Obviously the correlation exists. Its the strength of the correlation that’s being questioned, as well as the meaningfulness of the distinction. The idea that the gap between 2350 and 2380 or 2380 and 2400 has any sort of importance in and of itself is absurd.</p>