Statistical Question- Law School LSAT Range

(Applications Advice, Letters of Recommendation . . . )
User avatar
Lawquacious
Posts: 2037
Joined: Fri Oct 23, 2009 10:36 am

Statistical Question- Law School LSAT Range

Postby Lawquacious » Fri Jul 30, 2010 12:09 am

How is 25th-75th percentile band statistically derived? I understand that it represents the fact that the bulk (50%) of all students at that particular school fall within this range in terms of LSAT score (with a lower quartile of the students falling below this range and an upper quartile falling above this range), but how is it that some of the ranges are much wider than others? I guess you just literally take the middle 50% of the total number of scores and whatever the end points are (the extremes- lowest and highest individual numbers of this grouping) determines the range? Is it that simple?

Also, I have seen info that seems to indicate that the median LSAT score for a given school tends to be at the very upper end of this 25-75% percentile range. This has been confusing to me because I had the impression that this 25-75% percentile range represents a relatively evenly distributed score band (meaning that seeing a range of let's say 160-168 I would think that a 164 or 165 would be a pretty competitive score for that school because I assumed the approximate mean and/or median of the band fell around 164). However, if the band is calculated as described above and if the schools median LSAT is typically at the very upper end of the 25-75th band, then I think this could mean I have looked at this all wrong.

I am now thinking that comparing my score (or any given individual's score) to the school's median rather than where it falls in relation to the 25'th-75th range is likely to be much more predictive of acceptance at a given school. This is based on the idea that the 25-75th range does not represent a relatively even distribution of scores but rather just shows where the low and high points of the middle 50% of scores fall, with most of them probably actually falling toward the high part of the range given that overall median tends to be right around the high point.

Could someone who understands the basic statistical principles related to this weigh in?

User avatar
Knock
Posts: 5152
Joined: Wed Jun 10, 2009 3:09 pm

Re: Statistical Question- Law School LSAT Range

Postby Knock » Fri Jul 30, 2010 12:19 am

I think it's like, 25th = 25% of people had that LSAT or lower, 50th = 50% had that LSAT or lower, 75th = 75% had that LSAT or lower.

So like for Columbia, which has 25th at 170, median (50th) at 172, and 75th at 175, that just means 25% have a 170 or lower, 50% have a 172 or lower, 75% have 175 or lower.

Median is definitely going to be more useful when determining your likelihood of acceptance.

User avatar
Eugenie Danglars
Posts: 2353
Joined: Wed Jul 07, 2010 12:04 pm

Re: Statistical Question- Law School LSAT Range

Postby Eugenie Danglars » Fri Jul 30, 2010 12:55 am

To find a median layman's style: Line up all the numbers small to large, then pick up the one in the middle. That's your median.

Example:
12 15 19 20 23 25 32

Median here is 20.

The interesting example (for law school admissions at least) is this:

0 0 1 20 500 750 12034857829

Median is still 20.

25th and 75th percentiles are calculated similarly.

In theory, once you're above the 75th or below the 25th, your score shouldn't matter much. So if a school's 25th is 165, they shouldn't statistically prefer a 164 to a 150. Of course, in real life, it doesn't work this way, but it's cool, right?

User avatar
kazu
Posts: 1602
Joined: Sun Dec 06, 2009 9:35 pm

Re: Statistical Question- Law School LSAT Range

Postby kazu » Fri Jul 30, 2010 1:35 am

Eugenie Danglars wrote:To find a median layman's style: Line up all the numbers small to large, then pick up the one in the middle. That's your median.

Example:
12 15 19 20 23 25 32

Median here is 20.

The interesting example (for law school admissions at least) is this:

0 0 1 20 500 750 12034857829

Median is still 20.

25th and 75th percentiles are calculated similarly.

In theory, once you're above the 75th or below the 25th, your score shouldn't matter much. So if a school's 25th is 165, they shouldn't statistically prefer a 164 to a 150. Of course, in real life, it doesn't work this way, but it's cool, right?


To add to this, the reason law school LSATs medians are closer to the 75% is because their admittances become something like this:

0 5 10 15 20 70 71 72 73 74 75

Obviously this is a bit exaggerated, but above the median would be 70, the 25% would be 10 (the score of the person at the 25% line from the bottom), and the 75% would be 73.

User avatar
Lawquacious
Posts: 2037
Joined: Fri Oct 23, 2009 10:36 am

Re: Statistical Question- Law School LSAT Range

Postby Lawquacious » Fri Jul 30, 2010 2:18 am

Knockglock wrote:I think it's like, 25th = 25% of people had that LSAT or lower, 50th = 50% had that LSAT or lower, 75th = 75% had that LSAT or lower.

So like for Columbia, which has 25th at 170, median (50th) at 172, and 75th at 175, that just means 25% have a 170 or lower, 50% have a 172 or lower, 75% have 175 or lower.

Median is definitely going to be more useful when determining your likelihood of acceptance.


I think I understand the basics you mentioned in the first part of your post fairly clearly (I hope my post didn't give the impression that I do not). In any case, regarding your understanding that 50th percentile represents median, I am trying to understand (in part) how this assessment is not displaced given that the median score for each school actually falls very close to the 75th%percentile of LSAT score for that school. I realized that at the very least 25th/50th/75th percentile marks don't fall anywhere close to a 'flat' (this might not be the correct term to designate what I mean) distribution (e.g. U of MN has a 75th%tile of 168, but I believe their median is 167; SMU has a 75th of 165 with a median of 164).

Insofar as median does relate directly to 50th percentile (which I would like to confirm further, though it seems to make perfect sense), it just shows that 50th%tile is extremely skewed toward the 75th%tile in terms of actual number scores such that they are almost the same number. I suppose this could be possible given that a school will likely tend to take applicants who cluster fairly tightly around their median, and those above the median will probably for the most part be only slightly above it, thus restricting the quartile immediately above the 50th percentile to a very narrow range (the ones considerably above it are likely to go to a better school for one thing in many cases).

Also, the 50-75%tile could even include a portion of the LSAT scores consistent with the median number, since the same number score could represent both the median cut-off and a portion of the quartile immediately above the median (50th-75th) where the percentile cutoffs are based on actual ordered scores in a series rather than on a pre-set distribution.

The clustering around median, in some cases with a portion of the same individual median LSAT scores being counted for the quartile immediately below median and some for the quartile immediately above, and in all cases with those LSAT numerical scores immediately above the median number being counted toward the 50th-75th, could help to fill the 50'th-75th "quickly," and help explain why the 50th%tile is so close in terms of numerical LSAT score to the 75th. Conversely, the 50th percentile of an actual ordered number scale would not necessarily include all scores that are consistent with the top number of the 50th percentile (some of the identical number scores to the median could actually be included in the scores counted toward the 50-75th percentile instead), and the 50th percentile would also include numbers progressively less strongly clustered as they progress away from the median which would disproportionately expand the range of the 0-50th percentile more than the 50-75th.

User avatar
Lawquacious
Posts: 2037
Joined: Fri Oct 23, 2009 10:36 am

Re: Statistical Question- Law School LSAT Range

Postby Lawquacious » Fri Jul 30, 2010 2:23 am

kazu wrote:
Eugenie Danglars wrote:To find a median layman's style: Line up all the numbers small to large, then pick up the one in the middle. That's your median.

Example:
12 15 19 20 23 25 32

Median here is 20.

The interesting example (for law school admissions at least) is this:

0 0 1 20 500 750 12034857829

Median is still 20.

25th and 75th percentiles are calculated similarly.

In theory, once you're above the 75th or below the 25th, your score shouldn't matter much. So if a school's 25th is 165, they shouldn't statistically prefer a 164 to a 150. Of course, in real life, it doesn't work this way, but it's cool, right?


To add to this, the reason law school LSATs medians are closer to the 75% is because their admittances become something like this:

0 5 10 15 20 70 71 72 73 74 75

Obviously this is a bit exaggerated, but above the median would be 70, the 25% would be 10 (the score of the person at the 25% line from the bottom), and the 75% would be 73.


Thanks- that seems to be a nice illustration of the typical clustering effect.

WestOfTheRest
Posts: 1412
Joined: Mon Jul 07, 2008 11:10 pm

Re: Statistical Question- Law School LSAT Range

Postby WestOfTheRest » Fri Jul 30, 2010 4:14 am

If it makes you feel better, Harvard and Yale both appear to have a normal distribution based on their percentiles.

rundoxierun
Posts: 1893
Joined: Wed Dec 03, 2008 1:46 am

Re: Statistical Question- Law School LSAT Range

Postby rundoxierun » Fri Jul 30, 2010 12:58 pm

OP, think of 25th/median/75th numbers as "at least" numbers. For instance, a 25th percentile of 164 means AT LEAST 25% of the class scored at or below 164. It could be that 30% scored 164 or below, it could be that 30% scored exactly 164 and no one actually scored below.

09042014
Posts: 18282
Joined: Wed Oct 14, 2009 10:47 pm

Re: Statistical Question- Law School LSAT Range

Postby 09042014 » Fri Jul 30, 2010 1:04 pm

This isn't really a hard concept wtf is wrong with you.

User avatar
capitalacq
Posts: 639
Joined: Thu Feb 05, 2009 2:42 am

Re: Statistical Question- Law School LSAT Range

Postby capitalacq » Sat Jul 31, 2010 11:40 am

Desert Fox wrote:This isn't really a hard concept wtf is wrong with you.

liberal artsy majors :oops:

CanadianWolf
Posts: 10439
Joined: Wed Mar 24, 2010 4:54 pm

Re: Statistical Question- Law School LSAT Range

Postby CanadianWolf » Sat Jul 31, 2010 11:51 am

I understand your question & have had similiar thoughts on this matter as you. I don't know the answer in mathmatical terms, but will continue with the simplistic, and often correct, understanding expressed by the other posters here as well as by USNews.
The best way to evaluate your LSAT score is to utilize both the 25th & 75th percentile ranges as well as that law school's median LSAT score for matriculated students.
The concepts are quite simple, but it can be distressing when one sees a law school's mid-range at 166 to 172, but the median is 171.
P.S. I have read a few internet discussions in which a college or university's undergraduate figures show the matriculated first year students' median SAT as being above the 75th percentile figure for that school. Speculation runs from different year's data used to certain matriculated categories (e.g. URMs & development cases) were omitted from one set of data.
Last edited by CanadianWolf on Sat Jul 31, 2010 12:48 pm, edited 1 time in total.

Bumi
Posts: 947
Joined: Sun Jul 25, 2010 4:57 pm

Re: Statistical Question- Law School LSAT Range

Postby Bumi » Sat Jul 31, 2010 12:34 pm

CanadianWolf wrote:The concepts are quite simple, but it can be distressing when one sees a law school's mid-range at 166 to 172, but the median is 171.8.


A median of 171.8? This thread is classic.

270910
Posts: 2437
Joined: Thu May 21, 2009 9:51 pm

Re: Statistical Question- Law School LSAT Range

Postby 270910 » Sat Jul 31, 2010 12:39 pm

Desert Fox wrote:This isn't really a hard concept wtf is wrong with you.

CanadianWolf
Posts: 10439
Joined: Wed Mar 24, 2010 4:54 pm

Re: Statistical Question- Law School LSAT Range

Postby CanadianWolf » Sat Jul 31, 2010 12:49 pm

Corrected to 171. Thanks. I was trying to emphasize the OP's point of a cause of frustration for some.
OP: When you saw the numbers for a particular law school were they contained in the same article or from different sources even though addressing the same year ?




Return to “Law School Admissions Forum”

Who is online

Users browsing this forum: Bing [Bot], stratton_oakmont and 3 guests