Law School Stats Analysis Blog Forum

[californiauser]

Looking forward to Cornell and Penn. Keep up the good work! I hope you make some money off of this site!

[empyreanrrv]

BU does have a binding ED option, but it is only in its second year and there isn't much data up for it, so it may be hard to gauge. Might not even be worth trying to chart until a few years out, but thought I would mention it since the profile says it does not exist.

[elterrible78]

BU does have a binding ED option, but it is only in its second year and there isn't much data up for it, so it may be hard to gauge. Might not even be worth trying to chart until a few years out, but thought I would mention it since the profile says it does not exist.

I think you're probably right. I plan on updating this every application summer over the cycle, so we'll give it a few more years, but thanks for the heads up on this. I'll at least change the page to reflect this reality.

[elterrible78]

Hey, thanks to EVERYONE for the positive feedback!

Just wanted to let you know that I just put up the SLS profile, and it's definitely pretty interesting...confirms a lot of what you hear about SLS on these forums. I also included a "feedback" form on the right, for folks to let me know what kind of custom analysis they'd like to see (I want to compare schools above the Mason-Dixon line with schools below it in terms of URM boost is but one controversial topic that I think might be interesting to check out).

I'll continue updating as often as I can...it's all about balancing out getting through everything to post comparative analyses vs. putting new school profiles up.

Also, if anyone sees anything wrong (factual, typo, I have the wrong school name in the text of a school profile - I'm doing a lot of copy/paste, if you hadn't noticed), it'd be great if you could point it out so I can fix it.

Thanks!

[Ruluo]

Hey, thanks to EVERYONE for the positive feedback!

Just wanted to let you know that I just put up the SLS profile, and it's definitely pretty interesting...confirms a lot of what you hear about SLS on these forums. I also included a "feedback" form on the right, for folks to let me know what kind of custom analysis they'd like to see (I want to compare schools above the Mason-Dixon line with schools below it in terms of URM boost is but one controversial topic that I think might be interesting to check out).

I'll continue updating as often as I can...it's all about balancing out getting through everything to post comparative analyses vs. putting new school profiles up.

Also, if anyone sees anything wrong (factual, typo, I have the wrong school name in the text of a school profile - I'm doing a lot of copy/paste, if you hadn't noticed), it'd be great if you could point it out so I can fix it.

Thanks!

I don't think it's worth the time to attempt a Northern/Southern URM boost comparison (and not because it may be controversial--I don't think that should ever stand in the way of analysis). A big problem, however, is that the distribution of URM scores is lower, and if we assume normal distributions for the scores, we would think URMs would be underrepresented (relative to their % of test takers) at the higher end and overrepresented on the lower end. Imagine HYS, for example; they have fewer than 20 AA scoring above 168 and fewer than 100 above 164--those schools, along with the other most competitive schools have to reach far down the score ladder because of the dearth of high AA LSAT scores. Further down the score range, however, the distribution widens for AA, and the lower ranked schools do not have to drop their LSAT bar so low (relative to their medians) to accept a certain % of AA. Since most of the higher-ranked schools are in the North, they have to lower their LSAT standards more for those AA applicants (which would appear to be a bigger URM boost).

Unless you could control for those factors, I'm not sure how meaningful the analysis would be.

[Ruluo]

By the way elterrible, I appreciate the analysis.

Instead of attempting to determine static "boosts," have you considered including an interaction variable (if I understand correctly you're just running linear regressions)? The problem with the current boosts is that in reality they are very non-linear (a move from 179 to 180 probably won't have a significant impact on chances at HLS while a 171 vs a 172 might).

My stats knowledge is a little rusty, but if you were to set the dependent variable as the probability of acceptance and have LSAT, GPA, and an interaction variable as independent variables, you could give a more accurate indication of splitter effects (though frankly the results would be more difficult to interpret--this might lend itself better to a basic prediction calculator).

[elterrible78]

By the way elterrible, I appreciate the analysis.

Instead of attempting to determine static "boosts," have you considered including an interaction variable (if I understand correctly you're just running linear regressions)? The problem with the current boosts is that in reality they are very non-linear (a move from 179 to 180 probably won't have a significant impact on chances at HLS while a 171 vs a 172 might).

My stats knowledge is a little rusty, but if you were to set the dependent variable as the probability of acceptance and have LSAT, GPA, and an interaction variable as independent variables, you could give a more accurate indication of splitter effects (though frankly the results would be more difficult to interpret--this might lend itself better to a basic prediction calculator).

Thanks for these posts, and I definitely have chewed on everything you've brought up. The primary reason I haven't done the analysis the way you suggest is, for one (just as you state), interpreting results on interacted variables is just soooo difficult. Another reason is that for most of the schools in question, the number of observations on splitters and non-splitters is really low, which is why when it comes to splitters, non, and reverse-splitters, I prefer to just tabulate results and means and present it that way, along with a little heavily "caveated" analysis of what's in those tables. Running either ordered logistic regression, or even logistic regression, on that kind of data really tends to overblow things.

Again, thanks very much for the feedback, and I'm definitely open to any other suggestions.

[elterrible78]

I don't think it's worth the time to attempt a Northern/Southern URM boost comparison (and not because it may be controversial--I don't think that should ever stand in the way of analysis). A big problem, however, is that the distribution of URM scores is lower, and if we assume normal distributions for the scores, we would think URMs would be underrepresented (relative to their % of test takers) at the higher end and overrepresented on the lower end. Imagine HYS, for example; they have fewer than 20 AA scoring above 168 and fewer than 100 above 164--those schools, along with the other most competitive schools have to reach far down the score ladder because of the dearth of high AA LSAT scores. Further down the score range, however, the distribution widens for AA, and the lower ranked schools do not have to drop their LSAT bar so low (relative to their medians) to accept a certain % of AA. Since most of the higher-ranked schools are in the North, they have to lower their LSAT standards more for those AA applicants (which would appear to be a bigger URM boost).

Unless you could control for those factors, I'm not sure how meaningful the analysis would be.

This makes perfect sense, but do you think the specific problem that you are bringing up could be solved by simply controlling for each school's average USNWR ranking over the period we're drawing the data from?

EDIT: And even beyond any North/South comparison, it might be interesting just to see if URM boost varies predictably with school ranking, as you hypothesize (my first reaction is that it makes perfect sense and you're probably right).

[John_rizzy_rawls]

El, would it be at all possible to discern, or predict, what the URM medians (LSAT and GPA) of an individual school is from the data you have?

[elterrible78]

El, would it be at all possible to discern, or predict, what the URM medians (LSAT and GPA) of an individual school is from the data you have?

Pretty easy, actually! I'll take care of it next week when I get back.

[shifty_eyed]

Hey, long time no see! I just glanced through your site because I'm on my phone, but I think it will be very useful in providing data to reinforce or dispel TLS myths.

I would love to see a profile for Texas and don't want to wait until you get to the Ts! It has a large 65% (?) in-state requirement and became splitter friendly this cycle, so I think it would be interesting to look at.

Also, why do you think there is a female boost? Because there are fewer female applicants or because of the differences in LSAT distribution by gender?

[elterrible78]

Hey, long time no see! I just glanced through your site because I'm on my phone, but I think it will be very useful in providing data to reinforce or dispel TLS myths.

I would love to see a profile for Texas and don't want to wait until you get to the Ts! It has a large 65% (?) in-state requirement and became splitter friendly this cycle, so I think it would be interesting to look at.

Also, why do you think there is a female boost? Because there are fewer female applicants or because of the differences in LSAT distribution by gender?

Hey Shifty! I'm back, and I'll see if I can get this done sometime before I take off again. I'll do the medians thing for JRR (which could take a bit, actually) and your Texas profile before anything else, but I'll be traveling a lot over the next couple weeks, so I can't promise more than that. I'll let you know when it's done, though!

[nothingtosee]

BUMP

[xylocarp]

This is awesome! It really is fascinating to see these factors analyzed quantitatively.

Also, the design really isn't bad, and this is coming from a graphic designer.

[John_rizzy_rawls]

Hey, long time no see! I just glanced through your site because I'm on my phone, but I think it will be very useful in providing data to reinforce or dispel TLS myths.

I would love to see a profile for Texas and don't want to wait until you get to the Ts! It has a large 65% (?) in-state requirement and became splitter friendly this cycle, so I think it would be interesting to look at.

Also, why do you think there is a female boost? Because there are fewer female applicants or because of the differences in LSAT distribution by gender?

Hey Shifty! I'm back, and I'll see if I can get this done sometime before I take off again. I'll do the medians thing for JRR (which could take a bit, actually) and your Texas profile before anything else, but I'll be traveling a lot over the next couple weeks, so I can't promise more than that. I'll let you know when it's done, though!

Awesome, thanks El.

[elterrible78]

I'm on my way to move into my new place in Chicago right now. I've been all over the damn place over the past few weeks, so I haven't done jack with the site (Thailand and Cambodia, then home to Maine, now on my way to Chicago). I'll be leaving for a couple weeks in Miami, and during that time I'll definitely get on those special requests. Thanks for being patient, folks.

[jselson]

I'm on my way to move into my new place in Chicago right now. I've been all over the damn place over the past few weeks, so I haven't done jack with the site (Thailand and Cambodia, then home to Maine, now on my way to Chicago). I'll be leaving for a couple weeks in Miami, and during that time I'll definitely get on those special requests. Thanks for being patient, folks.

Lol, elterrible thanking US. Da man.

[elterrible78]

Shifty, the UT profile is up for you, and I am going to get to work on those URM/non-URM medians for JRR.

As always, any input, questions, or requests are appreciated!

[elterrible78]

Also, does anyone know where I could get medians for schools, short of going school-website by school-website looking for them?

[Lavitz]

You've got a typo in the Texas profile under the first model: "URM candidates also get quite a boost in terms of their chances at Stanford, being 20.4 times more likely to be accepted than an otherwise identical non-URM candidate."

Otherwise, looks good.

Also, does anyone know where I could get medians for schools, short of going school-website by school-website looking for them?

Well, there's the super spreadsheet, but Shadow stopped updating it, so it's incomplete. This thread also has a list of some of them in the OP.

You can also use the LSAC site. Find the school, click on "ABA Law school data" on the top right and the GPA/LSAT scores will be near the bottom of the page on the left.

[elterrible78]

You've got a typo in the Texas profile under the first model: "URM candidates also get quite a boost in terms of their chances at Stanford, being 20.4 times more likely to be accepted than an otherwise identical non-URM candidate."

Otherwise, looks good.

Also, does anyone know where I could get medians for schools, short of going school-website by school-website looking for them?

Well, there's the super spreadsheet, but Shadow stopped updating it, so it's incomplete. This thread also has a list of some of them in the OP.

You can also use the LSAC site. Find the school, click on "ABA Law school data" on the top right and the GPA/LSAT scores will be near the bottom of the page on the left.

Thanks, Lavitz...I can always count on you!

[nothingtosee]

Also, does anyone know where I could get medians for schools, short of going school-website by school-website looking for them?

http://www.lawschooltransparency.com/re ... s-Tracker/

25/50/75 for every school.

[Lavitz]

http://www.lawschooltransparency.com/re ... s-Tracker/

25/50/75 for every school.

It doesn't have class of 2015, but that's a great compilation.

[shifty_eyed]

Shifty, the UT profile is up for you, and I am going to get to work on those URM/non-URM medians for JRR.

As always, any input, questions, or requests are appreciated!

Thanks!

[elterrible78]

Just posted about URM and non-URM medians, JRR (and anyone else who's interested):

http://admissionsbythenumbers.blogspot. ... dians.html