How is the LSAT curved?

gambelda
Posts: 300
Joined: Tue Sep 28, 2010 11:44 am

How is the LSAT curved?

Postby gambelda » Tue Dec 07, 2010 11:47 am

I thought I knew the answer to this, but I guess I don't and I feel stupid now because of it. I always thought it was curved based on everyone numbers and then they took % of people who scored a set score in bands. i.e. 1.5% of the people who took the exam missed only 8 questions, so we'll make this years 170 curve a -8......if this is the case, how do preptests like 45 get released with a -13 curve? The consensus seems to be that PT 45 was really easy with a very leniant curve....but if it's easier across the board, wouldn't one expect it to have a harsher curve? Are curves predetermined?

tourdeforcex
Posts: 428
Joined: Fri Jun 11, 2010 2:19 pm

Re: How is the LSAT curved?

Postby tourdeforcex » Tue Dec 07, 2010 12:29 pm

i think you are correct in how you thought the LSAT is curved.

this consensus that you refer to... is this a TLS consensus? TLS is not representative. next, PT #45 could be easier relative to other tests and still have a lenient curve if everyone / most of the people who took the real deal in December 2004 had a hard time. so it could have been "harsh" for them. do you know what i mean? curves are not predetermined as far as i know.

gambelda
Posts: 300
Joined: Tue Sep 28, 2010 11:44 am

Re: How is the LSAT curved?

Postby gambelda » Tue Dec 07, 2010 12:36 pm

Yes, I'm basing it on a TLS consensus and I realized after posting that it could be that people who have practiced some of the newer exams in the 50's have gone back and did the 40's afterwards and found that the 50's really made the 40's easier. This seems to be the case with me as I have scored a 171 on PT 47 and 171 on PT 45 after doing a few 50's PT's.

My 171 on PT 45 had 4 really stupid mistakes in LR, partially because I was taking it on an airplane and I attribute those mistakes to the sort of cramped, noisier, high latitude ears failing to pop environment. So it seemed strange that I could jump substantially to say a 173-174 while the curve plummeted to a -13.

User avatar
4for44
Posts: 209
Joined: Sat Mar 13, 2010 4:05 pm

Re: How is the LSAT curved?

Postby 4for44 » Tue Dec 07, 2010 12:41 pm

gambelda wrote:I thought I knew the answer to this, but I guess I don't and I feel stupid now because of it. I always thought it was curved based on everyone numbers and then they took % of people who scored a set score in bands. i.e. 1.5% of the people who took the exam missed only 8 questions, so we'll make this years 170 curve a -8......if this is the case, how do preptests like 45 get released with a -13 curve? The consensus seems to be that PT 45 was really easy with a very leniant curve....but if it's easier across the board, wouldn't one expect it to have a harsher curve? Are curves predetermined?


From what I understand, curves are determined by performance of test takers when they were experimental... Otherwise LSAT mean would always be 150... which it is not... its gone up (i.e. the people taking the test later have a better understanding of the overall test due to new materials, overall learning curve). TLS correct me if I'm wrong

d34d9823
Posts: 1915
Joined: Wed Apr 14, 2010 2:52 pm

Re: How is the LSAT curved?

Postby d34d9823 » Tue Dec 07, 2010 12:42 pm

tourdeforcex wrote:curves are not predetermined as far as i know.

This is completely wrong.

The LSAT is not curved - it is equated. That is, the difficulty of the questions is determined by giving them in experimental sections of previous tests (this is why there is an experimental section). That known difficulty is then used to set the scoring key so that any given score correlates to the same level of performance across all tests. (The "margin of error" that LSAC releases with your test results is a description of the statistical certainty of this correlation.)

This is important because it adjusts for any demographic differences that may occur between test administrations due to economic fluctuations or preferences for a certain time of year. If the test was simply curved, one could play the system by determining which administration was likely to have the most lenient curve. Instead, all administrations have identically difficult tests, within LSAC's stated margin of error.

tourdeforcex
Posts: 428
Joined: Fri Jun 11, 2010 2:19 pm

Re: How is the LSAT curved?

Postby tourdeforcex » Tue Dec 07, 2010 12:43 pm

ahh thank you for that.

gambelda
Posts: 300
Joined: Tue Sep 28, 2010 11:44 am

Re: How is the LSAT curved?

Postby gambelda » Tue Dec 07, 2010 12:49 pm

d34dluk3 wrote:
tourdeforcex wrote:curves are not predetermined as far as i know.

This is completely wrong.

The LSAT is not curved - it is equated. That is, the difficulty of the questions is determined by giving them in experimental sections of previous tests (this is why there is an experimental section). That known difficulty is then used to set the scoring key so that any given score correlates to the same level of performance across all tests. (The "margin of error" that LSAC releases with your test results is a description of the statistical certainty of this correlation.)

This is important because it adjusts for any demographic differences that may occur between test administrations due to economic fluctuations or preferences for a certain time of year. If the test was simply curved, one could play the system by determining which administration was likely to have the most lenient curve. Instead, all administrations have identically difficult tests, within LSAC's stated margin of error.


So you're saying they sit down and assemble questions from experimental sections and say this section we wrote has X very difficult questions, Y moderat questions and Z easy questions....then based on thsoe numbers for the whole test they can equate a curve from a given formula they use?

Also, I'm just wondering, but how would curving it make the test biased so people could go to easier exams? couldnt any exam be significantly difficult but have a high curve when they determine that 1.5% missed 18 questions so we'll make the 170 curve a -18? i.e. how would one predict based off a curve which exam would be easiest?

User avatar
FlanAl
Posts: 1474
Joined: Fri Mar 12, 2010 4:53 pm

Re: How is the LSAT curved?

Postby FlanAl » Tue Dec 07, 2010 12:51 pm

If all administrations have equally difficult tests then why is the curve historically more lenient in december? Not questioning your response, I'm just still confused.

d34d9823
Posts: 1915
Joined: Wed Apr 14, 2010 2:52 pm

Re: How is the LSAT curved?

Postby d34d9823 » Tue Dec 07, 2010 1:24 pm

gambelda wrote:So you're saying they sit down and assemble questions from experimental sections and say this section we wrote has X very difficult questions, Y moderat questions and Z easy questions....then based on thsoe numbers for the whole test they can equate a curve from a given formula they use?

That's correct - the experimental section has two goals - to eliminate poorly written questions, and to determine the difficulty of questions for future use. I doubt it's as simple as categorizing them by difficulty. I would guess that they have a spreadsheet with a very accurate numerical description of each question's difficulty.

gambelda wrote:Also, I'm just wondering, but how would curving it make the test biased so people could go to easier exams? couldnt any exam be significantly difficult but have a high curve when they determine that 1.5% missed 18 questions so we'll make the 170 curve a -18? i.e. how would one predict based off a curve which exam would be easiest?

It is possible that June test takers are on average more competent than February test takers because of some underlying effect (e.g. taking in June shows preparedness and initiative vs. taking in February.) The pools of test takers are also heavily influenced by the state of the economy. Given this, if the test was simply curved, it might unfairly penalize one pool and reward the other pool of test takers.

d34d9823
Posts: 1915
Joined: Wed Apr 14, 2010 2:52 pm

Re: How is the LSAT curved?

Postby d34d9823 » Tue Dec 07, 2010 1:29 pm

FlanAl wrote:If all administrations have equally difficult tests then why is the curve historically more lenient in december? Not questioning your response, I'm just still confused.

The fact that the numerical 'curve' is different does not mean that equating is not occurring. A difficult test with a -14 'curve' and an easy test with a -9 can statistically equate to the same competence to score mapping.

I don't know whether LSAC intentionally varies the difficulty of the test or if it's just a function of the impossibility of producing a perfectly consistent 'curve'. I would guess that they vary it intentionally to enable people to choose whether they would prefer a hard test with a generous 'curve' or an easy test with a stingy 'curve', as many people (including myself) feel that they perform better at one extreme.

gambelda
Posts: 300
Joined: Tue Sep 28, 2010 11:44 am

Re: How is the LSAT curved?

Postby gambelda » Tue Dec 07, 2010 1:36 pm

Thanks for that information. It helps my understanding of the whole LSAC process a great deal. I believe I'm in the same boat as you, performing better at one extreme. I think I do better with difficult exams and higher curves.




Return to “LSAT Prep and Discussion Forum”

Who is online

Users browsing this forum: cctv, dontsaywhatyoumean, TakeItToTrial and 11 guests