rayiner wrote: Void wrote: rayiner wrote:
Void wrote:<Quoted Material>
I'm curious about the extent TLS thinks classes are fungible at similarly-ranked schools. I'm top 25%-ish at DNGC and I started wondering about this after I met a few students from UVA and a few from NYU and realized "you know, I don't think they're actually that much smarter than me."
Obviously, the classes at HYS are on average smarter than the classes at DNGC; just as obviously, it's clearly not the case that every Yale student is better than every Stanford student, who's better than every Harvard student, and on down the line. There are people at the top of the class at my school who I'm pretty sure would still be near the top at HYS. But, in those upper-middle to lower-middle ranges - how much of a difference do you think it would make if you took a top-quartile Georgetown student and dropped her at Columbia? Still top quartile? Top third? Median? Below median? My sense, admittedly based on almost no evidence, is that the performances wouldn't shift too much one way or the other.
</End Quoted Material>
Those students you met from UVA & NYU scored, what- maybe 5 more LSAT points than you did? That doesn't make them per se smarter. Sure, maybe different people have different LSAT ceilings, but how often has your LSAT experience had any affect on your class preparation whatsoever? How often do you even use the same part of your brain as you did for LSAT prep?
I concede that there are probably more gifted people at Yale than at a TTTT, but I don't think the divide even between those poles is anywhere near what your post suggests.
So the difference in LSAT median between Columbia and Georgetown is two points. With an R^2 of 0.5, LSAT would explain about one tenth of one standard deviation of grades. So someone who finished median at Columbia would, statistically, finish a hair above median at Georgetown.
Is this a joke?
No it's math. The point is that a 2-point LSAT difference will statistically result in barely any difference in grades.
Let me explain it another way. Let's assume an R^2 of 0.5 (a fairly high assumption). Now, say we compose a class with the full distribution of LSAT takers (not just law school admits, the whole range from 120-180). We give them all law school exams. The folks who got 170 on the exam are statistically only expected to finish one standard deviation above the median, or about top 15%.
Not to put too fine a point on it, but this math is both wrong and misleading.
The explanatory power of LSAT and GPA cannot be compared across schools, period. The fact of the matter is, the only group of people for which we have any even remotely comparable data at all would be transfer students, and do to significant selection bias and endogeneity problems that sample will never have enough students in it to yield anything like comparative data.
Georgetown takes students with high LSATs and high GPAs that Columbia does not take. Often, those GPAs at Georgetown mean less than the GPAs of comparable students at Columbia. Moreover, GPA/LSAT splitting (i.e. "gaming the medians") is much more common, and much more possible, the lower the law school's rank.
A student at Yale or Harvard is not only more likely to have done extremely well both on the LSAT and in undergrad, but is much more likely to have gone to a school from which their high GPA is an extremely strong indicator of ability. Yet, of course, because of the nature of the data that LSAC can collect, all that we can discern is that those with 4.2/180s do better than 3.8/173s more often by some amount (explained by the correlation coefficient). What we cannot say, at all
(and should not say) is that a 3.8/173 at Georgetown would have done just as well in the counter-factual situation, since that combination at Georgetown actually puts you near the tippy-top of their incoming GPA/LSATs, as opposed to the middle (moreover, that student's GPA is more likely to be from University of Phoenix, on average).
What this simple example should show (I hope) is that at best we can discern nothing about how students would do comparatively from the GPA/LSAT correlation data. Moreover, if anything, we should in fact conclude precisely the opposite of what you're saying. That is, if you think about it, given that 4.2/178s are doing better than 3.8/172s in comparable ratios to the ratios at lower ranked schools where the distribution is more likely comparing 4.2/165s and 2.1/175s we should be more apt to conclude, at least heuristically, that these data indicate that at HYS/CCN, on average, students would absolutely destroy students at schools ranked significantly lower.
Food for thought. This is Top-Law-Schools.com. You would expect more love for top law schools, no?