Doing Better at a "Worse" School Forum

[09042014]

What are you even talking about?

Obv trollin 2 the max

Not trolling. Not exactly advocating that an admitted student at Yale take his chances at Georgetown, just that all of you who think GPA and LSAT have no correlation to law school performance are, in fact, morons. Ya, the 75th and the 25th at any *single* school are going to be so close so as to make predictions on class rank difficult. But even then, I guarantee you there is some correlation. And when you're talking about an LSAT that exceeds a school's 90th percentile, you better damn believe that student has a decided advantage.

The problem is, even a Yale-worthy student can't count on beating out ~540 of his classmates at Georgetown in order to land one of the top five slots - there are just too many variables in play. Would I bet money on him being in the top 20%, or even top 15%? Possibly, but that's not an objectively better position than median at Yale.

The difference between Yale and Georgetown is pretty small in terms of student quality. Me and Rayiner did the math in a thread a couple years back using the LSAT correlation data.

Median at Yale wouldn't even be top quarter at Georgetown. Going off memory it'd likely be worse than top 40%.

[ScottRiqui]

Median at Yale wouldn't even be top quarter at Georgetown. Going off memory it'd likely be worse than top 40%.

Yeah, I know I was being generous with the top 20%/top 15% thing, but the 3.9/178 given in the example were pretty strong numbers, even for Yale (median GPA, and two points over the 75th LSAT). But even assuming a top 20% outcome, I still don't think it would objectively put the Georgetown student in a better position than a median-ish Yale student.

[arklaw13]

I love how the OP posts real data showing there is a correlation between LSAT/GPA and first year grades and numerous posters respond that there is no correlation between LSAT/GPA and first year grades.

ScottRiqui's response was the correct one.

The OP's correlation for upga was almost as strong as the LSAT correlation. You're telling me you believe that raw ugpa, unweighted for what school or program you attended, whether there was significant grade inflation/deflation, has any predictive value for law school exam performance? The only thing I would trust ugpa to measure across schools is how consistently drunk you were during undergrad.

[ScottRiqui]

I am under my school's 25th LSAT percentile, and I am in the top 4%. THERE IS LITTLE TO NO CORRELATION WITH LAW SCHOOLS EXAMS AND LSAT SCORES

One anecdotal story isn't statistically significant.

You should retake.

I'm not going to discount his case - a data point is a data point. But it goes back to the ecological fallacy that cron1834 was talking about earlier; just as you can't use aggregate statistics about an entire population to make predictions about a single member of that population, you can't deny the presence of aggregate trends just because one member of the population doesn't conform. I'd bet that if you looked at the LSAT/GPA and grades of his entire class, the correlation would be there.

[Tiago Splitter]

I love how the OP posts real data showing there is a correlation between LSAT/GPA and first year grades and numerous posters respond that there is no correlation between LSAT/GPA and first year grades.

ScottRiqui's response was the correct one.

The OP's correlation for upga was almost as strong as the LSAT correlation. You're telling me you believe that raw ugpa, unweighted for what school or program you attended, whether there was significant grade inflation/deflation, has any predictive value for law school exam performance? The only thing I would trust ugpa to measure across schools is how consistently drunk you were during undergrad.

I don't know. Talk to the dudes who did the study. All I know is that their study found a correlation.

The better argument is that the correlation is not a strong enough factor on which to make a decision, not that there is no correlation.

[ScottRiqui]

I love how the OP posts real data showing there is a correlation between LSAT/GPA and first year grades and numerous posters respond that there is no correlation between LSAT/GPA and first year grades.

ScottRiqui's response was the correct one.

The OP's correlation for upga was almost as strong as the LSAT correlation. You're telling me you believe that raw ugpa, unweighted for what school or program you attended, whether there was significant grade inflation/deflation, has any predictive value for law school exam performance? The only thing I would trust ugpa to measure across schools is how consistently drunk you were during undergrad.

I don't know. Talk to the dudes who did the study. All I know is that their study found a correlation.

The better argument is that the correlation is not a strong enough factor on which to make a decision, not that there is no correlation.

Exactly. Here's an example of several data sets with different correlations:



The top-middle plot shows a positive correlation of 0.3 (which is pretty close to the median correlation between uGPA and 1L grades). You can see that there's a trend there, but it's not strong at all - there are a lot of points that don't fall anywhere near the prediction line. The 0.7 plot is better, but still not great as a predictive tool, even if we had pre-law school data that would give us a 0.7 correlation with law school performance.

[Hipster but Athletic]

I want to see stats on rigor of undergrad institution & ls performance.

[kartelite]

Someone accepted Yale with a 3.9/178 would have a fantastic chance of being top 10% at Georgetown, you'd have to be crazy not to think that. They've proven that their ability to absorb course material and perform on a very difficult test is much, much higher than that of the "typical" student at a school like Georgetown.

There are also factors that cause significant downward bias to the reported correlations between LSAT/UGPA and law school grades. In the table linked below (won't let me post images), we have law school grades showing negative correlation with both LSAT and UGPA at two example schools. Does this mean that someone with higher grades or LSAT demonstrates lower ability to be a successful law student? No - someone with a higher LSAT/UGPA is more likely to be at School A, competing with higher quality classmates. And someone at school B with a high LSAT (Student 3) is in general going to have a lower GPA than someone with a lower LSAT (Student 2). Furthermore, even combining UGPA and LSAT, a student with a 3.75/172 who gets into Yale has probably demonstrated exceptional talent in other areas (or had a very challenging UG school/major, or aced their classes in grad school) and should be expected to perform fairly similarly to the 3.9/178 student.

LSAT correlation example

[Tiago Splitter]

Someone accepted Yale with a 3.9/178 would have a fantastic chance of being top 10% at Georgetown, you'd have to be crazy not to think that.

It's funny because you'd actually have to be crazy to think anyone has a "fantastic" chance of finishing in the top 10% at Georgetown.

[ScottRiqui]

Someone accepted Yale with a 3.9/178 would have a fantastic chance of being top 10% at Georgetown, you'd have to be crazy not to think that.

It's funny because you'd actually have to be crazy to think anyone has a "fantastic" chance of finishing in the top 10% at Georgetown.

Never mind the fact that the original claim wasn't top 10% at Georgetown, it was top five students, which is top <1%.

[aboutmydaylight]

An r = .36 means that roughly only 13% of the variance in first year grades can be explained by the LSAT score of the student alone. That's a pretty terrible amount, and it really shouldn't give any individual confidence that they could outperform their grades significantly at a lower ranked school, even though it happens on average.

[jn7]

I would bet a lot of money that if they ran the regression against scores only above the low 160s you would see no correlation at all. I really think that a correlation shows up only because of a "hard work" factor that distinguishes people with 140 vs. actual scores you needed to study for. Just my two cents and I'm no statistician.

[aboutmydaylight]

I would bet a lot of money that if they ran the regression against scores only above the low 160s you would see no correlation at all. I really think that a correlation shows up only because of a "hard work" factor that distinguishes people with 140 vs. actual scores you needed to study for. Just my two cents and I'm no statistician.

Iirc the correlation coefficients are within each school (which is why there isn't just one, but multiple). So yea there would be some correlation because at the top schools everyone has good LSAT scores and there is still a correlation. I will take a guess that the correlation is much lower at top ranked schools whereas the .5s or whatever are at really low ranked schools.

[woosah]

The better argument is that the correlation is not a strong enough factor on which to make a decision, not that there is no correlation.

I agree with this. I think it's a fairly straightforward argument:


1. If there is a factor which contributes to the chance that you finish in a higher percentage of a law school's class, you should consider that factor when making your decision to go to that law school.

2. Your LSAT/GPA relative to a law school's average LSAT/GPA is a factor which contributes to the chance that you finish in a higher percentage of that law school's class.

Conclusion: You should consider your LSAT/GPA relative to a law school's average LSAT/GPA when making your decision to go to that school.


And I think (2) is shown by the fact that there is a correlation between LSAT/GPA and average 1L grades. The more interesting debate becomes: "Is that 'factor' significant/useful enough to warrant any substantive consideration?"


ETA: This thought experiment may be way off base, but consider two schools that are EXACTLY the same in every feature. One has an average LSAT/GPA of x, the other has an average LSAT/GPA of 0.9x. Given what we know about the correlation studies, it seems to me to be smarter to pick the one with the average LSAT/GPA of 0.9x. (If this thought experiment works, it seems to strengthen the argument I gave above.)

[Clearly]

The better argument is that the correlation is not a strong enough factor on which to make a decision, not that there is no correlation.

I agree with this. I think it's a fairly straightforward argument:


1. If there is a factor which contributes to the chance that you finish in a higher percentage of a law school's class, you should consider that factor when making your decision to go to that law school.

2. Your LSAT/GPA relative to a law school's average LSAT/GPA is a factor which contributes to the chance that you finish in a higher percentage of that law school's class.

Conclusion: You should consider your LSAT/GPA relative to a law school's average LSAT/GPA when making your decision to go to that school.


And I think (2) is shown by the fact that there is a correlation between LSAT/GPA and average 1L grades. The more interesting debate becomes: "Is that 'factor' significant/useful enough to warrant any substantive consideration?"

The correlation is no where near strong enough that it should have a role in decisions, plus even if it was, the correlation vs rankings should balance out anyway. Top ten % at wustl vs median at duke vs bottom third at Harvard etc.

[woosah]

even if it was, the correlation vs rankings should balance out anyway. Top ten % at wustl vs median at duke vs bottom third at Harvard etc.

That's a good point. Ok. I'm done.

[ScottRiqui]

ETA: This thought experiment may be way off base, but consider two schools that are EXACTLY the same in every feature. One has an average LSAT/GPA of x, the other has an average LSAT/GPA of 0.9x. Given what we know about the correlation studies, it seems to me to be smarter to pick the one with the average LSAT/GPA of 0.9x. (If this thought experiment works, it seems to strengthen the argument I gave above.)

If the two schools were *exactly* the same in every regard (including job placement), then sure - go be the big(ger) fish in the small(er) pond. But as you go down in the rankings, the odds of getting one of the traditional "desirable outcomes" (i.e. biglaw or A3 clerkship) fall off a lot faster than the median LSAT/uGPA.

[jbagelboy]

Regardless of the merits of the median/LS class rank correlation, I think the argument that student quality diminishes at a much slower rate than employment prospects carries the day with this question 99/100 times. As has been mentioned, to whatever extent a 166/3.6 differs in student quality from a 169/3.75 (probably more than zero ceteris paribus, but not very much), a Duke student has a comparatively much better shot at gainful employment than a George Washington student.

The relevant considerations of this nature that would give one cause to choose the "worse" school are 1) price & debt, and 2)location & target market.

Also, USNWR is a joke. So this whole debate is largely moot.

[jbagelboy]

ETA: This thought experiment may be way off base, but consider two schools that are EXACTLY the same in every feature. One has an average LSAT/GPA of x, the other has an average LSAT/GPA of 0.9x. Given what we know about the correlation studies, it seems to me to be smarter to pick the one with the average LSAT/GPA of 0.9x. (If this thought experiment works, it seems to strengthen the argument I gave above.)

If this were true, it would mean everyone choosing between Harvard and Stanford (comparable employment, prestige, ranking, ect.) should go to Stanford, because both its medians are lower year after year, so by your logic, it would be easier competition. Empirically this is entirely unfounded.

[PepperJack]

Splitters seem more likely to stop caring once they have a job locked up, and plummet. So after 3 years, there'll be some correlation once we all fall enough.

[kartelite]

ETA: This thought experiment may be way off base, but consider two schools that are EXACTLY the same in every feature. One has an average LSAT/GPA of x, the other has an average LSAT/GPA of 0.9x. Given what we know about the correlation studies, it seems to me to be smarter to pick the one with the average LSAT/GPA of 0.9x. (If this thought experiment works, it seems to strengthen the argument I gave above.)

If this were true, it would mean everyone choosing between Harvard and Stanford (comparable employment, prestige, ranking, ect.) should go to Stanford, because both its medians are lower year after year, so by your logic, it would be easier competition. Empirically this is entirely unfounded.

The reason it doesn't work is that Stanford uses a more holistic admissions process, so it ends up admitting applicants who have demonstrated ability in ways other than LSAT/GPA. A 3.82/168 with a master's in quantitative finance from Princeton who spent two years founding a successful non-profit will likely find more love from Stanford than from Harvard. A K-JD with mediocre softs and a 3.95/176, not so much.

The problem with the correlation data is that they are done by school. Students at any given school should be fairly close in terms of "ability," and someone with a 178 at Duke is likely to have a worse UGPA and other credentials than the Duke student who scored 164. This could actually lead to the sort of situation described by Simpson's Paradox, as conceivably UGPA could be negatively correlated with law school grades at a given school if most people there with high UGPAs had low LSATs and people with low UGPAs had high LSATs (and LSAT ended up being the more important factor in law school exam performance).

It would be more meaningful to compare LSAT to bar passage rates, since that would get rid of the school bias. Obviously, coefficients from a multiple regression model would be much more instructive for the original question (and a logistic model would be great for modeling bar passage).

[patogordo]

here's a puzzle for you: let's say you're choosing between 3 schools, with LSAT medians of 160, 160, and 170. you have a 165 lsat. but you retake and get a 175. you visit school #3 and the dean reveals a goat in her office. do you switch schools?

[kartelite]

here's a puzzle for you: let's say you're choosing between 3 schools, with LSAT medians of 160, 160, and 170. you have a 165 lsat. but you retake and get a 175. you visit school #3 and the dean reveals a goat in her office. do you switch schools?

When someone reveals a goat to you, you always switch.

[Archangel]

here's a puzzle for you: let's say you're choosing between 3 schools, with LSAT medians of 160, 160, and 170. you have a 165 lsat. but you retake and get a 175. you visit school #3 and the dean reveals a goat in her office. do you switch schools?

When someone reveals a goat to you, you always switch.

http://www.philosophyexperiments.com/mo ... fault.aspx

[Hutz_and_Goodman]

here's a puzzle for you: let's say you're choosing between 3 schools, with LSAT medians of 160, 160, and 170. you have a 165 lsat. but you retake and get a 175. you visit school #3 and the dean reveals a goat in her office. do you switch schools?

Lol