<p>Ali v. Frazier, Karpov v. Kasparov, and now … siserune v. fabrizio - a modern classic! First of all, siserune, you are an awesome writer, way up there in that rarified intellectual meta-world of rhetoric. But fabrizio does have a way of pointing out nettlesome inconsistencies and uncomfortable facts, in his own workmanlike manner. You say “is” may or may not be “is”, he says “A or not A, which is it?” Awesome! So far, it’s a hard-fought draw!</p>
<p>^If you’re looking for the KO, fabrizio just conceded the “Asian correlates negatively with talent given score” thing. The other dispute was whether the “Asians work harder” meme is bogus, for which he has always pleaded No Contest on the substantive question (of whether there is any identifiable difference between Asian academic behavior and painless rational agency).</p>
<p>
</p>
<p>If anyone other than fabrizio thinks I’ve been vague on some still unaddressed substantive point and wants it clarified, just say what that is and I’ll see what I can do.</p>
<p>
</p>
<p>If anyone other than fabrizio thinks I’ve failed to disambiguate some important point, just quote the statement and I’ll see what I can do.</p>
<p>This offer applies to currently existing CC posters with enough postings to show that they are not sock puppets.</p>
<p><a href=“CanuckGuy:”>quote</a></p>
<p>I do have a question for you, siserune. Your theory concerning how SAT scores over-estimate the ability of Asian applicants is interesting. I would expand your theory and argue that the scores certainly over-estimate the ability of the URMs. If the elites compensate for this with the Asians, why are they not do so for the URMs as well?
[/quote]
</p>
<p>The arguments aren’t limited to SAT (recall the old discussions of similar phenomena with math competitions) or Asians. The mechanism of underperformance is different and less clear for URM in general, but similar in principle. First of all, hard working blacks or striving Martians face the same type of underperformance forecast as hard working Asian; strivers get creamed relative to credentials eventually, no matter where they come from. Beyond that, the variables are different. For Asians differential effort and differential family support are the obvious things that are not sustainable in college, while for URM or the poor it might often be standing out from a weaker crowd, or something as simple as luck on the SAT. In a merciless academic selection you should probably discount URM, the poor and most other disadvantaged groups. This is undesirable and/or impossible for social reasons, but that is what a pure predictive selection would look like. To some degree, “making it” from a diminished environment indicates greater talent at any given score, but in the long run the cumulative disadvantages of money problems, family troubles, and dysfunctional social networks probably should cause a bigger Dickensian discount than the “diamond in the rough” bonus.</p>
<p>I think this kind of regime would be quite brutal, but it is hard to deny the logic from the point of view of meritocratic selection. Also, there is some value to “cognitive diversity” from the pure performance point of view so even if your only goal is to have the richest graduates or win a technology race, some amount of affirmative-action like process to ensure some diversity might be important, but the justification for the current levels of AA is not academic in nature.</p>
<p>^I hear you loud and clear. The fly in the ointment is that not all URMs are disadvantaged and not all white and Asians are advantaged. Giving privileged URMs (Ogbus study in Shaker Heights comes immediately to mind) such a huge advantage over poor whites and Asians is problematic. Giving it to descendants of slaves is one thing, but also giving it to well educated African immigrants whose ancestors were probably responsible for the slavery is inflammatory in more ways than one.</p>
<p>Based on the Duke data, and the following graph,</p>
<p>[CARPE</a> DIEM: Chart of the Day: GRE Scores By Academic Field](<a href=“http://mjperry.blogspot.com/2008/11/chart-of-day-gre-scores-by-academic.html]CARPE”>CARPE DIEM: Chart of the Day: GRE Scores By Academic Field)</p>
<p>I would predict few URMs would end up in a STEM field in Duke. In other words, you can create artificial diversity through admission, but you can not create real diversity in the classroom. Your comment about SAT and women is interesting, and is probably the reason why there are not more women in STEM fields as well.</p>
<p>This reminds me of the last time you had a debate with fabrizio. Many posters jumped in to support you until you mentioned SAT scores and women. Then they jumped off just as fast, if not faster. Priceless.</p>
<p>I think I know who the East Asian genetic supremacist is. You are not talking about the guy who graduated from Cal Tech, and made millions selling a start-up, are you? I think you made a few comments on his blog at one time, if my memory is correct, and stopped for some reason. I think his critics need you there. Too many of them simply dont have the firepower to go up against a theoretical physicist. I know I certainly dont. They are about as bright as they come.</p>
<p>As I see it from the Great White North anyway.</p>
<p>Several talented kids I know locally ended up in University of Houston or another local college as the last choice possible. They assumed they were so brilliant that they should only apply to Harvard or Stanford and nowhere else (reeks of sheer laziness if you ask me).</p>
<p>Several hardworking kids I know who actually put in the time to fill their applications, ensured all paperwork is complete and did not leave anything to chance ended up at Harvards and Stanfords.</p>
<p>High school kids who think and know they have talent usually get burned because of arrogance. One of them had over 235 in PSAT, blew off most of the senior year classesonly applied to Harvard and a couple of other schools and scrambled to find something to go to in April. So usually in the end, even if talent is attributed to the kids who end up in the top schools, most of them tend to be really hard working irrespective of their race.</p>
<p>
</p>
<p>That’s right, and it is one of the reasons why AFTER the selection into the top schools, the later and more difficult selections (grad school fellowships, university math competitions, tenure appointments, etc) will sort those mostly really hard-working students by factors other than work ethic.</p>
<p>For later selections that are easier than the admission the opposite will occur and the hard worker proportion will increase. An example is graduation rates.</p>
<p>
</p>
<p>Sure: your “meritocratic discounting” argument is a joke. If you assume that SAT scores are increasing functions of talent and effort, then yes, holding a score constant, talent and effort are perfectly negatively correlated by construction (i.e. the relationship is mechanical).</p>
<p>Apparently, that the relationship is mechanical wasn’t of any concern to siserune. Nor did his introduction of a selection bias by holding the score constant ring any alarms. Nope, despite (seemingly) KNOWING full well that the perfect negative correlation is spurious due to Berkson’s Paradox / Fallacy, he CONTINUES to argue that a legitimate policy prescription of the spurious negative correlation is that Asians’ scores should be “meritocratically discounted” since being Asian is positively correlated with effort and thus, thanks to selection bias, negatively correlated with talent given a fixed score.</p>
<p>How sordid. If he didn’t know about Berkson’s Paradox, then it was fine for him to make the “meritocratic discounting” argument. But that he KNEW such conditional correlations are spurious and yet STILL made the argument is puzzling. It suggests that he knows that his arguments are invalid; he just wants to see if anyone can catch him. To his credit, I could not, as I did not know about the name or existence of Berkson’s Paradox until yesterday.</p>
<p>
</p>
<p>Is “mechanical” your word for what is usually called “deterministic”? The argument, as I said before, does not rely on having a deterministic relationship.</p>
<p>Let’s say that Score is an increasing function of several variables, such as Talent, Academic Preparation, and Luck on the day of the test (presumably not a deterministic quantity). If growing up in an Asian household tends to raise one’s level of preparation, all else being equal, then when interpreting a particular score (eg., 750) from an Asian candidate your estimate of Talent and of Luck should be lower than for a non-Asian with the same information. </p>
<p>Good news: Asians with high SAT scores are less likely to have achieved them through random chance than non-Asians with the same score! Is this model with a random Luck term, “mechanical”? Because it exhibits the same negative correlations you are calling spurious.</p>
<p>
</p>
<p>Why should it be a matter for concern if something is “mechanical” (and what does that word mean exactly, in ordinary mathematical terms)?</p>
<p>
</p>
<p>No, it merely relies on knowing the name “Berkson’s Paradox” without understanding the paradox’s implication. Whether you say that score is now a function of three variables does not change that all you did was give another example of “Berkson’s Paradox” and then make a policy prescription without realizing that your prescription is based on a biased finding / a fallacy / a paradox.</p>
<p>While sis was kind enough to share two links defining “Berkson’s Paradox,” that he ever so conveniently ignoring the name of the paradox itself should make it obvious to everyone that he has no incentive to actually explain the paradox / the fallacy / the bias. So let’s consider the paradox in greater detail, following [Roberts</a> et al. (1978)](<a href=“http://www.epidemiology.ch/history/PDF%20bg/Roberts%20RS%20et%20al%201978%20an%20empirical%20demonstration%20of%20berkson’s%20bias.pdf]Roberts”>What is epidemiology? | Lernzeit.de).</p>
<p>Berkson was a physician scientist who warned about the dangers of interpreting results from seemingly well-controlled medical studies, which often studied an important question in epidemiology: the association between the exposure to a suspected causal factor of disease and the presence of a disease. These studies often employed the case-control approach, where the level of exposure to the hypothesized cause is determined in a treatment group (the cases) and a control group. The control group is drawn from patients admitted to a hospital with a disease not thought to have an association with the suspected causal factor </p>
<p>A reading of the introduction in Roberts et al. shows a detailed demonstration of Berkson’s bias / fallacy / paradox. The question is whether hypertension is a risk factor for skin cancer. Suppose that the prevalence of hypertension, skin cancer, and accidental bone fractures is INDEPENDENT **in the general population<a href=“i.e.%20unconditionally”>/b</a>; the occurrence of one in a person does not affect the likelihood of the same person having one of the others. Thus, by assumption, there is no link between hypertension and skin cancer in the general population.</p>
<p>The example shows that if hospitalization rates of people with hypertension, skin cancer, and accidental bone fractures differ, then **in the hospital<a href=“i.e.%20conditionally”>/b</a>, there will be an association between hypertension and skin cancer, even though in the general population, there was no association!</p>
<p>Thus, in the world of the Roberts et al. example, any claim that there exists a link between hypertension and skin cancer is SPURIOUS. The sad part here is that sis knows a concept (“Berkson’s Paradox”) but has no idea what it means. sis’s “meritocratic discounting” is a sham argument based on a lack of understanding of the implication of Berkson’s bias / fallacy / paradox.</p>
<p>
</p>
<p>Does the instance of the argument that I posted in #368 still involve “mechanical correlations”? If asking two or three times for a definition of what that phrase means was a bit excessive on my part earlier, I apologize. Just typing the word YES or NO would clear things up splendidly.</p>
<p>
</p>
<p>Your “argument” merely relies on knowing the name “Berkson’s Paradox” without understanding its implication. Unconditionally, there’s no relationship between Asian and talent, but with our assumptions of SAT score determinants, yes, conditioning on a given score, Asian and talent are negatively related, even though in the general population, there was no such relationship. </p>
<p>But that negative conditional dependence is spurious; it doesn’t exist in the general population (i.e. unconditionally). It’ll hold true for any given score, but it’ll never be true in the general population. So again, your “meritocratic discounting” is a complete sham.</p>
<p><a href=“CanuckGuy:”>quote</a></p>
<p>This reminds me of the last time you had a debate with fabrizio. Many posters jumped in to support you until you mentioned SAT scores and women. Then they jumped off just as fast, if not faster. Priceless.
[/quote]
</p>
<p>One person was OK with it, for the rest I obtained some data tracking the progression from PSAT to SAT, GRE, GMAT and LSAT showing that the SAT writing advantage for women is eliminated as men go to college and are forced to read books, while men maintain an unyielding advantage on the upper range of the verbal test. The thread was locked and it is tedious to collate the numbers for posting, but maybe one day if it becomes topical. I did want to answer your question about the political motivation, though: the writing section was added to the PSAT as settlement of a lawsuit that the original Math+Verbal test was denying women money from the National Merit Scholarship program.</p>
<p>
</p>
<p>It should be noted that he is not a racist whatsoever, only an EAGS with a microphone and a genetics lab. </p>
<p>
</p>
<p>Good catch! I did spank him on the blog about the Espenshade statistical studies, after he quoted them (as unnamed “studies that show…”) in a widely circulated news article much discussed in CC. That article is now a central hub in the recursive self-citation bubble of Asian admission discrimination reports, where each new article references the ones before as the proof, the old reports crystallizing into the status of hard facts under the weight of more recent texts. Universities have their own PR levers but the guerillas are winning this round of the propaganda war.</p>
<p>The article was written by guy’s Yale friend (the one who has been misrepresenting her qualifications on admissions) who appears to be his “in” to the media. After a short exchange he wrote on the blog that “if Espenshade’s results are correct” he would have to retract his comments to the Boston Globe, something that has not happened, even on the blog. On the other hand, if Espenshade’s results are NOT correct, then most of the citation bubble explodes, so there is an incentive to avoid the issue entirely.</p>
<p>
</p>
<p>FYI, the word “Berkson” never appeared on this site until your post #359 a day or two ago. Are you saying that a posted argument with no mention of this paradox, and working in a different mathematical setup, invisibly “relies” on it anyway, with the invisible part being visible to you? And that you were able to determine my degree of knowledge or understanding of the paradox although — as far as the visible spectrum is concerned — I never actually wrote anything about it? </p>
<p>This is the second time in as many days that you explicitly claim personal psychic powers. I don’t have any specific reason to doubt that you have those skills. But as just a slight concession to those of us without the paranormal abilities, I was wondering if you could also provide the list of errors in the NON-invisible part of my argument, highlighting some of the words that are wrong and can also be seen on the computer screen. If you can explain why they are wrong, so much the better. Many thanks.</p>
<p>
</p>
<p>Well, gee, I wonder why. Maybe it’s because until “a day or two ago,” I didn’t know what Berkson’s Paradox was. Are we forgetting that you’re the one who indirectly introduced the concept to the thread?</p>
<p>
</p>
<p>Now, you may retort, in your inimitable fashion, “There goes fab again with his fab-rications. It will be evident to all the readers that I did not once mention the word ‘Berkson’ in either of the examples I gave.”</p>
<p>Uh, yeah, that’s what I said in #359: “you ever so conveniently ignored the name of the concept you’re using with your conditional correlation assertion: Berkson’s paradox.”</p>
<p>What did sis excise from Dr. Murphy’s web page? “In statistics, this is known as Berkson’s paradox, or “selection bias”. For a dramatic example of this effect,…” sis HAD to have read the sentence that preceded the one from which he began quoting, yet he feigns ignorance as to the term “Berkson’s paradox.”</p>
<p>And what did sis excise from page 17 of Causality by Dr. Judea Pearl? “However, this corresponds to a general pattern of causal relationships: observations on a common consequence of two independent causes tend to render those causes dependent, because information about one of the causes tends to make the other more or less likely, given that the consequence has occurred. This pattern is known as selection bias or Berkson’s paradox in the statistical literature… For example, if the admission criteria…” Again, sis HAD to have read the two sentences that preceded the one from which he began quoting, yet again, he pretends that he has no idea what “Berkson’s paradox” is…</p>
<p>
</p>
<p>…or that his two “arguments” are merely instances of Berkson’s bias / fallacy / paradox!</p>
<p>PS: FYI, the name “Dornbusch” never appeared on this site from you until earlier this week. Search is a fun tool, isn’t it, sis?</p>
<p>
</p>
<p>A copy of the argument, all one sentence of it, is appended at the bottom of this post. That’s in case you would like to quote errors in the argument instead of just placing the word “argument” between quotes.</p>
<p>
</p>
<p>Unconditionally, you can assume any relationship you like, such as Asians being super-talented geniuses, or the opposite. It does not affect the argument.</p>
<p>
</p>
<p>Right. But why would you utilize the relationship for the general population when evaluating individuals whose scores are known (e.g., for a college admission)? </p>
<p>
</p>
<p>Words like “spurious” and “paradox” and “bias” applied to this situation have no content beyond the statement that the conditional and unconditional distributions give different signs for the correlations. They have no bearing on which distribution is the correct one to use for the problem of prediction from given information.</p>
<p>The good old argument, once again: </p>
<p>
</p>
<p>
</p>
<p>Of course you didn’t. If you did, you’d be exposing your “meritocratic discounting” canard for the sham that it is. I called you out on this in #369, which you cleverly excised as “deletia”: “While sis was kind enough to share two links defining “Berkson’s Paradox,” that he ever so conveniently ignoring the name of the paradox itself should make it obvious to everyone that he has no incentive to actually explain the paradox / the fallacy / the bias.”</p>
<p>
</p>
<p>I already did in “deletia.” In #369, I cited [Roberts</a> et al. (1978)](<a href=“http://www.epidemiology.ch/history/PDF%20bg/Roberts%20RS%20et%20al%201978%20an%20empirical%20demonstration%20of%20berkson’s%20bias.pdf]Roberts”>What is epidemiology? | Lernzeit.de) and outlined their example of Berkson’s bias / fallacy / paradox. They easily show that if hospitalization rates of people with hypertension, skin cancer, and accidental bone fractures differ, then in the hospital (i.e. conditionally), there will be an association between hypertension and skin cancer, even though in the general population, there was no association, by assumption.</p>
<p>The implication of that example is that any “finding” of a link between hypertension and skin cancer is SPURIOUS. And what does spurious mean, sis? “Apparently but not actually valid.” Mathematically, in the hospital, a link between hypertension and skin cancer exists. But by assumption, there was no link between the two in the general population. The link only exists because of SELECTION BIAS. That’s Berkson’s Paradox: people can easily find relationships when in reality there is none due to seemingly little things like selection bias.</p>
<p>The tl;dr version for you, my dear sis, is that your “meritocratic discounting” argument is a joke, or to paraphrase you, the kind that gets doctoral students in sociology “defrocked.”</p>
<p>
</p>
<p>Congratulations! You just repeated, verbatim, your “argument” for “meritocratic discounting” based on a MECHANICAL RELATIONSHIP. Oops, sis doesn’t like that term. So why don’t we demonstrate why it’s mechanical?</p>
<p>Suppose the following linear relationship between score, talent, preparation, and luck, for tractability:</p>
<p>score = a<em>talent + b</em>prep + cluck; a, b, c > 0*</p>
<p>Suppose that we hold the score constant at 750.</p>
<p>750 = a<em>talent + b</em>prep + cluck*</p>
<p>And hey, purely for purposes of demonstration, let’s make the example more numerical:</p>
<p>750 = 20<em>talent + 30</em>prep + 10luck*</p>
<p>Let’s consider i, a non-Asian candidate with talent 20, prep 10, and luck 5 who earns a 750. Let’s also consider j, an Asian candidate whose prep is 20 but whose talent and luck are unknown. Suppose that we assume j has the same luck as i.</p>
<p>750 = 20<em>talent + 30</em>20 + 105*
talent = 5 < 20</p>
<p><em>Gasp</em>! Given a score of 750, j is far less talented than is i! OMG, sis has proved that at any given score, Asians are likely to be less talented than are non-Asians! I smell a Nobel!</p>
<p>…except anyone can see the problem with that “argument.” The relationship is mechanical. It’s an equality, so to maintain equality given a fixed score, if something goes up, something else has to go down. In our case, if prep goes up, talent and/or luck have to go down to maintain the equality.</p>
<p>That’s all. There’s nothing special about sis’s “argument” as quoted above. His finding of a negative correlation between Asian and talent given a fixed score is nothing but an example of Berkson’s paradox. The sad part is, sis can cite examples of the paradox without realizing that his “argument” is merely another example.</p>
<p>So I didn’t smell a Nobel for sis; I smelled an (Ig) Nobel.</p>
<p>
</p>
<p>How is that a problem with the argument? The class of models expressible in “mechanical” form is extremely general. Of those, the ones that are monotonic (ie, the ones to which this simple argument applies) are also a very broad category. Simplicity and broad applicability are positives, so what exactly is not to your liking?</p>
<p>
</p>
<p>The time difference between me linking to documents with interesting new words (e.g., Berkson paradox, “selection bias”, “spurious correlation”) and the onset of the ongoing word-salad of half baked terminology was 30 minutes. </p>
<p>Not that I doubt the unusual mental powers seen in this thread, but speaking just as someone who has actually studied that material for periods longer than 48 hours I thought I might, I dunno, alert you to the possibility that maybe a bit more contemplation is in order before inscribing your new-skimmed expertise on these words into a lecture series stored on a permanenty archived internet forum. I mean, who knows how it could look if you say something really stupid, like repetitively shrieking at people for understanding what turns out to be very basic math.</p>
<p>
</p>
<p>You’re confused as to what the concept is and whether my argument “relies” on it.</p>
<p>
</p>
<p>Berkson’s paradox is analogous (or a special case) but is NOT a good way to explain the matter at hand, which is why I excluded it from the two excerpts, why I have been reluctant to answer your postings in terms of Berkson’s setup, and may be part of why you are totally confused in your #369 and other postings on the same theme. If you scream and shriek for a discussion in “Berkson” terms I may eventually do that just to stop the cluttering of the thread, but trust me that you will reach enlightenment faster by disregarding it. Or not; pursue your intuitions and see where that goes…</p>
<p>
</p>
<p>uh, I profess the near-certainty that you will only confuse yourself fixating on Berkson’s paradox. Certainly I know what it is and the argument I described is obviously from the same general family of ideas that goes by many names. </p>
<p>
</p>
<p>That’s right. And I posted information from Dornbusch here for the last several years. You didn’t know that?</p>
<p>I agree with MisterK. Watching the sis and fab show is one of the great pleasures of being on CC; I learned a lot from both of them. There are a few others here I would put in the same category as well.</p>
<p>Never heard of Berkson’s Paradox before. Sounds like it is a close relative of Simpson’s Paradox, but not the Abilene Paradox. Am I correct?</p>