Law School Stats Analysis Blog Forum

[ScottRiqui]

Makes sense - thanks!

[littleaaron]

i wish i wasn't so awful at statistics.

But it seems to me that a traditional splitter having a higher lsat score (or having a higher GPA to go with a high lsat score) would be better.

For example, a 2.9/177 would be better than a 2.9/176, and a 2.8/175 would be worse than a 2.9/175. Just from a pure numbers standpoint.

Am I wrong here, and why? It seems that since each entering person has their gpa and lsat calculated, there are two considerations for a splitter: how much the lsat will boost medians and how much their gpa will hurt medians.

IN the first case above (why a 2.9/177 > 2.9/176), they take the same GPA hit, but in exchange for the SAME gpa median hit they get a slight increase in their lsat median with the first applicant.
Similarly, in the second situation above (2.8/175 < 2.9/175), they get the same lsat median boost, but with the second applicant they take LESS of a gpa median hit.

So, I can't image why, from a pure numbers standpoint, increasing your lsat score or increasing your GPA wouldn't confer an advantage, regardless of whether you're the lowest GPA in the class or the highest lsat. It seems that, since both are being calculated, in ALL cases, it's better (advantageous) to have both be as high as possible.

Of course, I'm dumb, and I see three possible errors here: First, I might just have no idea how they calculate medians. I assume they take all of them and then calculate. OF course, it's possible that they do some weird system of only taking the interquartile range or something, thus removing the 25/75th percentiles, which would be the smart move as far as statistics is concerned, in which case my entire argument is moot. But being wrong is how you learn, right?

It's also possible that the differences I'm describing are so negligible that they don't matter. I kind of doubt that's the case, as it seems that if the lsat boost gained by taking a splitter is worth it, so too is the gpa hit something to consider. But again, stats are weird and i don't get them as well as i should.

or maybe schools just care way more about their lsat median? I realize that lsat medians are considered 'more' important, but i assume that the gpa median is important enough to not want it to drop if possible.

There was one more problem I saw with this, but i forget now because it took me so long to type all this other junk.

Of course, I'd be so happy of someone were to tell me that all people over the 75th lsat and under the gpa 25th have an equal chance, and explain why.... because that's definitely me. But i'm also a sub 3.0, which i assume makes a significant difference.

[elterrible78]

Am I wrong here, and why? It seems that since each entering person has their gpa and lsat calculated, there are two considerations for a splitter: how much the lsat will boost medians and how much their gpa will hurt medians.

IN the first case above (why a 2.9/177 > 2.9/176), they take the same GPA hit, but in exchange for the SAME gpa median hit they get a slight increase in their lsat median with the first applicant.
Similarly, in the second situation above (2.8/175 < 2.9/175), they get the same lsat median boost, but with the second applicant they take LESS of a gpa median hit.

I think you may be thinking of the mean (average) rather than the median. Yes?

Of course, I'd be so happy of someone were to tell me that all people over the 75th lsat and under the gpa 25th have an equal chance, and explain why.... because that's definitely me. But i'm also a sub 3.0, which i assume makes a significant difference.

I don't think this is actually the case at all. You'll run across milder forms of this argument on TLS, but it's really hyperbole. From what I have seen so far, for splitters the LSAT really doesn't matter much, whereas the GPA does (and the reverse is true of high-GPA, low-LSAT candidates).

[littleaaron]

Holy. F*ck.

Yes, you're right... I've been thinking of this as means for a really, really long time now. It just made sense to me that they would care about the average, and never really read the word "median" and thought about the implications. I just saw a "stats term" and assumed it was the average.

Well, shoot Thanks for clearing that up for me.

But does this still hold for how they calculate 25/75th percentiles, or do they do that in some strange way, too?

So, then, it would seem that the only advantage to being over 75th (other than saying "we got this guy who's over 75th") is to boost your 75th percentiles... but for schools, all they would actually care about if medians were the only stat available would be if one was over median, right? Assuming that you could work your numbers in order to raise or maintain your median?

i.e. for splitters, the only number that would matter would be that you were over the target median, if medians were all that mattered?

thanks!!!!

[elterrible78]

But does this still hold for how they calculate 25/75th percentiles, or do they do that in some strange way, too?

Same thing...just regular old 25ths and 75ths.

So, then, it would seem that the only advantage to being over 75th (other than saying "we got this guy who's over 75th") is to boost your 75th percentiles... but for schools, all they would actually care about if medians were the only stat available would be if one was over median, right? Assuming that you could work your numbers in order to raise or maintain your median?

i.e. for splitters, the only number that would matter would be that you were over the target median, if medians were all that mattered?

thanks!!!!

This is certainly true to a degree and, like I said, some of the more cynical TLSers think that this is precisely how it happens, believing admissions to be simply a numbers game, and no more. But the fact of the matter is that, beyond gaming USNWR rankings (which totally happens, don't get me wrong), those numbers are representative of ability, potential, work ethic, raw talent, etc. So no, it's not a matter of you being over the median on your LSAT and they don't give a rat's ass about your GPA. Theoretically, in a pure numbers-gaming scenario this would be true, but admissions is not a pure numbers-gaming scenario.

[elterrible78]

Okay, splitter post is up. Go ahead and check it out if you're interested. There are charts with colors and everything!

http://www.admissionsbythenumbers.com

[elterrible78]

I hate to be that guy and bump my own thread, but I since the splitter issue comes up so often, it's justified.

[ScottRiqui]

Thanks for putting that together! Is there a way to anchor the column headings so that they remain visible when you scroll through the list of schools?

[elterrible78]

Thanks for putting that together! Is there a way to anchor the column headings so that they remain visible when you scroll through the list of schools?

I'll see if I can't figure something out.

In the meantime, does anyone know when NU started doing their full-schollies for ED accepts thing?

[Lavitz]

In the meantime, does anyone know when NU started doing their full-schollies for ED accepts thing?

2011-2012 cycle.

[elterrible78]

In the meantime, does anyone know when NU started doing their full-schollies for ED accepts thing?

2011-2012 cycle.

Thanks man.

[nothingtosee]

Could I put in a U Chicago request?

[elterrible78]

Could I put in a U Chicago request?

Will try to get on it today. Thanks!

[you'rethemannowdawg]

Awesome work, elterrible. This is really good stuff.

One thing that jumped out at me was that it seems like characterizations of splitter-friendliness based off of distance from the mean yields strange results. For example, NYU admitted almost 90% of splitters, yet is characterized "neutral-unfriendly." WUSTL and W&L admitted almost 100% of splitters and are called "neutral-friendly."

These three schools all seem pretty "splitter-friendly" to me. Is this a result of all schools placing heavier emphasis on LSAT than GPA?

[elterrible78]

Awesome work, elterrible. This is really good stuff.

One thing that jumped out at me was that it seems like characterizations of splitter-friendliness based off of distance from the mean yields strange results. For example, NYU admitted almost 90% of splitters, yet is characterized "neutral-unfriendly." WUSTL and W&L admitted almost 100% of splitters and are called "neutral-friendly."

These three schools all seem pretty "splitter-friendly" to me. Is this a result of all schools placing heavier emphasis on LSAT than GPA?

Thanks!

What you mention is part of it, but if you look, I think a big part of the story with, for example, NYU vs. WUSTL is that WUSTL is willing to dip wayyyyy lower on the GPA scale to snag the higher LSATs. If you look at the mean accepted GPA for non-splitters vs. splitters at NYU, the difference is only 0.36 (3.79 vs. 3.43). At WUSTL, it's 0.62 (3.6 vs 2.98). I think this matters when we are evaluating how splitter-friendly a school is. In order to even do analysis like this, we have to define what a splitter is, and how we do that marks a cutoff. Because of this, there are different levels of splitterdom, if you will.

So, say a school's cutoff GPA to be considered a splitter is 3.5 and below, and splitters have a 95% success rate at that school. How meaningful would that 95% be if, say, the average GPA for splitters was 3.49? Not all that meaningful, right?

This is all, to some degree, just as much art as science, but does that explanation make sense?

That said, it's a good point - the distance from the mean thing. It's fine for comparing among schools (by my measure, WUSTL is more splitter-friendly than NYU, bottom line), but in an absolute sense, it might lack a little. I'm definitely open to any suggestions of other criteria for categorization, provided that it's not something like "if a school admits X% of splitters, it's splitter-friendly." Even just tweaking the formula to give more/less weight to certain factors, as long as it makes sense and more people seem to support it.

[you'rethemannowdawg]

Awesome work, elterrible. This is really good stuff.

One thing that jumped out at me was that it seems like characterizations of splitter-friendliness based off of distance from the mean yields strange results. For example, NYU admitted almost 90% of splitters, yet is characterized "neutral-unfriendly." WUSTL and W&L admitted almost 100% of splitters and are called "neutral-friendly."

These three schools all seem pretty "splitter-friendly" to me. Is this a result of all schools placing heavier emphasis on LSAT than GPA?

Thanks!

What you mention is part of it, but if you look, I think a big part of the story with, for example, NYU vs. WUSTL is that WUSTL is willing to dip wayyyyy lower on the GPA scale to snag the higher LSATs. If you look at the mean accepted GPA for non-splitters vs. splitters at NYU, the difference is only 0.36 (3.79 vs. 3.43). At WUSTL, it's 0.62 (3.6 vs 2.98). I think this matters when we are evaluating how splitter-friendly a school is. In order to even do analysis like this, we have to define what a splitter is, and how we do that marks a cutoff. Because of this, there are different levels of splitterdom, if you will.

So, say a school's cutoff GPA to be considered a splitter is 3.5 and below, and splitters have a 95% success rate at that school. How meaningful would that 95% be if, say, the average GPA for splitters was 3.49? Not all that meaningful, right?

This is all, to some degree, just as much art as science, but does that explanation make sense?

That said, it's a good point - the distance from the mean thing. It's fine for comparing among schools (by my measure, WUSTL is more splitter-friendly than NYU, bottom line), but in an absolute sense, it might lack a little. I'm definitely open to any suggestions of other criteria for categorization, provided that it's not something like "if a school admits X% of splitters, it's splitter-friendly." Even just tweaking the formula to give more/less weight to certain factors, as long as it makes sense and more people seem to support it.

Oh I see. It does make sense to characterize schools as splitter-friendly based not just on the amount of splitters admitted, but also the numerical quality of the splitters. It's helpful to see the analysis of the T14 related to each other that you did farther down on the blog, because a splitter for the T14 is often not a splitter for other schools.

[elterrible78]

Oh I see. It does make sense to characterize schools as splitter-friendly based not just on the amount of splitters admitted, but also the numerical quality of the splitters. It's helpful to see the analysis of the T14 related to each other that you did farther down on the blog, because a splitter for the T14 is often not a splitter for other schools.

Cool, glad that made sense! As far as your last observation, though...I've coded each school's applicants as either splitters, reverse-splitters, or non-splitters based on that particular school's numbers for the cycle that the applicant applied, so someone who would be a splitter at one school but not a splitter at another consequently gets coded as a splitter at that school, and not at the other.

[whippersnappery]

This is really awesome. Would it be possible to see GPA/LSAT averages for women and URM, much in the same way that you have them for splitters/non-splitters?

[elterrible78]

This is really awesome. Would it be possible to see GPA/LSAT averages for women and URM, much in the same way that you have them for splitters/non-splitters?

That's no problem at all. Just let me know which schools you'd like to see them for, or if you just want all of them, let me know (that'll take a bit, though)!

[whippersnappery]

This is really awesome. Would it be possible to see GPA/LSAT averages for women and URM, much in the same way that you have them for splitters/non-splitters?

That's no problem at all. Just let me know which schools you'd like to see them for, or if you just want all of them, let me know (that'll take a bit, though)!

Thanks! I would be most interested in seeing how that works for T6.

[elterrible78]

This is really awesome. Would it be possible to see GPA/LSAT averages for women and URM, much in the same way that you have them for splitters/non-splitters?

That's no problem at all. Just let me know which schools you'd like to see them for, or if you just want all of them, let me know (that'll take a bit, though)!

Thanks! I would be most interested in seeing how that works for T6.

Here you go!

Average LSAT/GPA for women and URM accepted applicants are:

Yale: Women (174 / 3.89) & URM (170.1 / 3.77)
Harvard: Women (173.3 / 3.86) & URM (169.2 / 3.72)
Stanford: Women (172.2 / 3.87) & URM (168.9 / 3.70)
Chicago: Women (171.8 / 3.77) & URM (167.4 / 3.69)
Columbia: Women (173 / 3.77) & URM (167.8 / 3.65)
NYU: Women (172.2 / 3.76) & URM (167.5 / 3.65)

Hope that's what you were looking for!

[Humbert Humbert]

This is really awesome. Would it be possible to see GPA/LSAT averages for women and URM, much in the same way that you have them for splitters/non-splitters?

That's no problem at all. Just let me know which schools you'd like to see them for, or if you just want all of them, let me know (that'll take a bit, though)!

Thanks! I would be most interested in seeing how that works for T6.

Here you go!

Average LSAT/GPA for women and URM accepted applicants are:

Yale: Women (174 / 3.89) & URM (170.1 / 3.77)
Harvard: Women (173.3 / 3.86) & URM (169.2 / 3.72)
Stanford: Women (172.2 / 3.87) & URM (168.9 / 3.70)
Chicago: Women (171.8 / 3.77) & URM (167.4 / 3.69)
Columbia: Women (173 / 3.77) & URM (167.8 / 3.65)
NYU: Women (172.2 / 3.76) & URM (167.5 / 3.65)

Hope that's what you were looking for!

This is interesting. Out of curiosity, can you run the #s for these schools for men as well?

[elterrible78]

Here you go!

Average LSAT/GPA for women and URM accepted applicants are:

Yale: Women (174 / 3.89) & URM (170.1 / 3.77)
Harvard: Women (173.3 / 3.86) & URM (169.2 / 3.72)
Stanford: Women (172.2 / 3.87) & URM (168.9 / 3.70)
Chicago: Women (171.8 / 3.77) & URM (167.4 / 3.69)
Columbia: Women (173 / 3.77) & URM (167.8 / 3.65)
NYU: Women (172.2 / 3.76) & URM (167.5 / 3.65)

Hope that's what you were looking for!

This is interesting. Out of curiosity, can you run the #s for these schools for men as well?

Sure. For all men (URM & non-URM included, just as I did for the women):

Yale: Men (174.6 / 3.88)
Harvard: Men (174.2 / 3.86)
Stanford: Men (172.8 / 3.86)
Chicago: Men (172.5 / 3.78)
Columbia: Men (173.7 / 3.78)
NYU: Men (172.9 / 3.77)

The close match of the GPAs between men and women is pretty interesting.

[24671]

This is really awesome. Would it be possible to see GPA/LSAT averages for women and URM, much in the same way that you have them for splitters/non-splitters?

That's no problem at all. Just let me know which schools you'd like to see them for, or if you just want all of them, let me know (that'll take a bit, though)!

Thanks! I would be most interested in seeing how that works for T6.

Here you go!

Average LSAT/GPA for women and URM accepted applicants are:

Yale: Women (174 / 3.89) & URM (170.1 / 3.77)
Harvard: Women (173.3 / 3.86) & URM (169.2 / 3.72)
Stanford: Women (172.2 / 3.87) & URM (168.9 / 3.70)
Chicago: Women (171.8 / 3.77) & URM (167.4 / 3.69)
Columbia: Women (173 / 3.77) & URM (167.8 / 3.65)
NYU: Women (172.2 / 3.76) & URM (167.5 / 3.65)

Hope that's what you were looking for!

Thanks so much for this would you be able to find the medians based off urm type?? I know it's a lot to ask, either way, really appreciate it.

[elterrible78]

Thanks so much for this would you be able to find the medians based off urm type?? I know it's a lot to ask, either way, really appreciate it.

As much as I'd love to be able to, limitations with the data makes it impossible. Sorry!