Most lenient curve ever?

twintipping_bumps
Posts: 196
Joined: Tue Jun 02, 2009 11:14 am

Most lenient curve ever?

Postby twintipping_bumps » Sun Feb 07, 2010 4:11 pm

Does anyone know the most lenient curve ever? I have done some searches but really found nothing.

Alot of people seem to agree the LGs from February were very difficult in terms of being able to do them in time. When I started prep, LG was my weakest area by far. I got pretty good at them: lots of -2 and -3s and one -0. I got a -2 on the section in an actual sitting during September. But I had a tough time yesterday. The first one took like 10 minutes, and I never really understood the third.

I am wondering if we are in the midst of a trend back toward harder games and more lenient curves. Could December be the start of this, or just an outlier in a sea of -10s and -11s?

Also frustrating is that Feb was only 100 questions. Since I am more like a 86 or 87 credited responses taker during my last few months regardless of the curve, I prefer that 101st question for another shot at a credited response.

Will only 100 questions make a -14 repeat impossible. Anyone think the curve might be more lenient, i.e. there have now been enough administrations in this economy to attract less prepared takers and since LSAC still needs to keep a certain percentage at, say, 165 or above, it now needs to be more lenient with the curve because there are only a finite number of people that can do extraordinarily well but since massive amounts of people are now sitting they still need to be able to fill the final 15 points in the score range with a certain percentage of total test takers?

Anyone think Feb is more lenient than -14?

User avatar
monkeyboy
Posts: 181
Joined: Sat Dec 05, 2009 9:11 pm

Re: Most lenient curve ever?

Postby monkeyboy » Sun Feb 07, 2010 4:19 pm

twintipping_bumps wrote:Does anyone know the most lenient curve ever? I have done some searches but really found nothing.

Alot of people seem to agree the LGs from February were very difficult in terms of being able to do them in time. When I started prep, LG was my weakest area by far. I got pretty good at them: lots of -2 and -3s and one -0. I got a -2 on the section in an actual sitting during September. But I had a tough time yesterday. The first one took like 10 minutes, and I never really understood the third.

I am wondering if we are in the midst of a trend back toward harder games and more lenient curves. Could December be the start of this, or just an outlier in a sea of -10s and -11s?

Also frustrating is that Feb was only 100 questions. Since I am more like a 86 or 87 credited responses taker during my last few months regardless of the curve, I prefer that 101st question for another shot at a credited response.

Will only 100 questions make a -14 repeat impossible. Anyone think the curve might be more lenient, i.e. there have now been enough administrations in this economy to attract less prepared takers and since LSAC still needs to keep a certain percentage at, say, 165 or above, it now needs to be more lenient with the curve because there are only a finite number of people that can do extraordinarily well but since massive amounts of people are now sitting they still need to be able to fill the final 15 points in the score range with a certain percentage of total test takers?

Anyone think Feb is more lenient than -14?


Probably not more lenient. If we're lucky, it will be the same. Probably -11 or -12, but people were saying -11 after December, so we just can't know for sure at this point.

eros
Posts: 58
Joined: Sat Sep 26, 2009 8:18 pm

Re: Most lenient curve ever?

Postby eros » Sun Feb 07, 2010 4:20 pm

more lenient than -14, i dunno about that. but i could definitely see it being -14 for sure. i took the december test as well and everything besides reading comp was more difficult this time around.

User avatar
Kohinoor
Posts: 2756
Joined: Sat Oct 25, 2008 5:51 pm

Re: Most lenient curve ever?

Postby Kohinoor » Sun Feb 07, 2010 4:20 pm

twintipping_bumps wrote:Does anyone know the most lenient curve ever? I have done some searches but really found nothing.

Alot of people seem to agree the LGs from February were very difficult in terms of being able to do them in time. When I started prep, LG was my weakest area by far. I got pretty good at them: lots of -2 and -3s and one -0. I got a -2 on the section in an actual sitting during September. But I had a tough time yesterday. The first one took like 10 minutes, and I never really understood the third.

I am wondering if we are in the midst of a trend back toward harder games and more lenient curves. Could December be the start of this, or just an outlier in a sea of -10s and -11s?

Also frustrating is that Feb was only 100 questions. Since I am more like a 86 or 87 credited responses taker during my last few months regardless of the curve, I prefer that 101st question for another shot at a credited response.

Will only 100 questions make a -14 repeat impossible. Anyone think the curve might be more lenient, i.e. there have now been enough administrations in this economy to attract less prepared takers and since LSAC still needs to keep a certain percentage at, say, 165 or above, it now needs to be more lenient with the curve because there are only a finite number of people that can do extraordinarily well but since massive amounts of people are now sitting they still need to be able to fill the final 15 points in the score range with a certain percentage of total test takers?

Anyone think Feb is more lenient than -14?

Don't they determine the curve prior to administration of the exam? If so, what you suggest is impossible since the process is not reactive.

User avatar
vanwinkle
Posts: 9740
Joined: Sun Dec 21, 2008 3:02 am

Re: Most lenient curve ever?

Postby vanwinkle » Sun Feb 07, 2010 4:22 pm

Kohinoor wrote:Don't they determine the curve prior to administration of the exam? If so, what you suggest is impossible since the process is not reactive.

This. Unusually poor performance would be taken into account in calculating future curves, but the curve is generated in advance based on the difficulty of the test compared to the tests administered in the last 3 years.

User avatar
s0ph1e2007
Posts: 1054
Joined: Sat Oct 10, 2009 10:37 pm

Re: Most lenient curve ever?

Postby s0ph1e2007 » Sun Feb 07, 2010 4:24 pm

vanwinkle wrote:
Kohinoor wrote:Don't they determine the curve prior to administration of the exam? If so, what you suggest is impossible since the process is not reactive.

This. Unusually poor performance would be taken into account in calculating future curves, but the curve is generated in advance based on the difficulty of the test compared to the tests administered in the last 3 years.


I thought it was based on how people did on past experimental sections?

twintipping_bumps
Posts: 196
Joined: Tue Jun 02, 2009 11:14 am

Re: Most lenient curve ever?

Postby twintipping_bumps » Sun Feb 07, 2010 4:27 pm

I was thinking more in terms of: if the last four administrations or so have seen a significant influx of takers due to the economy, and these new takers are, as some have suggested, generally less prepared or not 100% committed to attending Law School, then wouldn't that mean that there has been time for these takers to take experimental sections and contribute to the calculations of how they will be curved when inserted into actual tests?

twintipping_bumps
Posts: 196
Joined: Tue Jun 02, 2009 11:14 am

Re: Most lenient curve ever?

Postby twintipping_bumps » Sun Feb 07, 2010 4:29 pm

s0ph1e2007 wrote:
vanwinkle wrote:
Kohinoor wrote:Don't they determine the curve prior to administration of the exam? If so, what you suggest is impossible since the process is not reactive.

This. Unusually poor performance would be taken into account in calculating future curves, but the curve is generated in advance based on the difficulty of the test compared to the tests administered in the last 3 years.


I thought it was based on how people did on past experimental sections?


That was my understanding as well. As I have just posted above, one thing I am wondering is if the December curve was an indication of a new trend that will cope with the influx of test takers.

User avatar
monkeyboy
Posts: 181
Joined: Sat Dec 05, 2009 9:11 pm

Re: Most lenient curve ever?

Postby monkeyboy » Sun Feb 07, 2010 4:29 pm

s0ph1e2007 wrote:
vanwinkle wrote:
Kohinoor wrote:Don't they determine the curve prior to administration of the exam? If so, what you suggest is impossible since the process is not reactive.

This. Unusually poor performance would be taken into account in calculating future curves, but the curve is generated in advance based on the difficulty of the test compared to the tests administered in the last 3 years.


I thought it was based on how people did on past experimental sections?


I think it's based on how people have done on these sections before.

I think people were scoring better than expected in the 40-49 exam range, so tests have been getting progressively more difficult as they try to balance out the distribution properly. More and more people are studying, there are more prep companies, and people are performing better as a result. I would bet LSAC is trying to figure out how to make this test truly do what it is supposed to in terms of the distribution.
Last edited by monkeyboy on Sun Feb 07, 2010 4:33 pm, edited 1 time in total.

februaryftw
Posts: 170
Joined: Mon Dec 07, 2009 12:01 pm

Re: Most lenient curve ever?

Postby februaryftw » Sun Feb 07, 2010 4:33 pm

monkeyboy wrote:
s0ph1e2007 wrote:
vanwinkle wrote:
Kohinoor wrote:Don't they determine the curve prior to administration of the exam? If so, what you suggest is impossible since the process is not reactive.

This. Unusually poor performance would be taken into account in calculating future curves, but the curve is generated in advance based on the difficulty of the test compared to the tests administered in the last 3 years.


I thought it was based on how people did on past experimental sections?


I think it's based on how people have done on these sections before.

I think people were scoring better than expected in the 40-49 PT range, so tests have been getting progressively more difficult as they try to balance out the distribution properly. More and more people are studying, there are more prep companies, and people are performing better as a result. I would bet LSAC is trying to figure out how to make this test truly do what it is supposed to.


Yes, it is likely that LSAC doesn't want a -8 curve, hence the uptick in games and rc difficulty; if they kept giving out games like those in the low 40s, it throws out some of the distinctions at the top that you want to preserve (e.g. no 179s, etc.).

User avatar
monkeyboy
Posts: 181
Joined: Sat Dec 05, 2009 9:11 pm

Re: Most lenient curve ever?

Postby monkeyboy » Sun Feb 07, 2010 4:39 pm

Seriously, reading comp and games were both easier in the 40s. LR questions were still somewhat difficult, but the rest of the test became very learnable. Some tweaking is probably going on. I think the good news is that one can have a not so great section (based solely on December and the feedback from yesterday) and still score near PT average, as I did in December.

User avatar
vanwinkle
Posts: 9740
Joined: Sun Dec 21, 2008 3:02 am

Re: Most lenient curve ever?

Postby vanwinkle » Sun Feb 07, 2010 4:50 pm

s0ph1e2007 wrote:
vanwinkle wrote:
Kohinoor wrote:Don't they determine the curve prior to administration of the exam? If so, what you suggest is impossible since the process is not reactive.

This. Unusually poor performance would be taken into account in calculating future curves, but the curve is generated in advance based on the difficulty of the test compared to the tests administered in the last 3 years.


I thought it was based on how people did on past experimental sections?

They use experimental sections to test out new questions and new question types, to gauge their difficulty and also to ensure they're being missed/gotten by the correct people. (For example it'd be odd if there were a question that most of the 170+ people got wrong and a bunch of people in the 150-160 range to right.) But in order to do that they have to measure the answers in the experimental section to people's performance on the real test.

Then they weigh all that information to generate a curve that, based on their information, will compare easily to the difficulty of tests over the prior 3 years. So a 170 on this test should correlate to a 170 on the last 3 years' worth of tests in terms of difficulty.

I hope that makes it clearer.

twintipping_bumps
Posts: 196
Joined: Tue Jun 02, 2009 11:14 am

Re: Most lenient curve ever?

Postby twintipping_bumps » Sun Feb 07, 2010 4:57 pm

So do you guys think there is that much of a lag time between experimental sections and their appearances on real tests, i.e. sections from the 40s only beginning to appear now. Some people yesterday were saying they saw sections they had recently had as experimental sections. That does seem real quick, though. If I had a RC I had had as an experimental, I don't think that would be right. I would remember alot of it.

But I guess I haven't worked out my thoughts on the curve process either. I think the performance on the experimental sections has to be the main determinant, but there must be some other formula to account for amount of test takers and the quality of performance that is statistically likely for a given number of people. I mean, if you have 30,000 people, you may get 1% of people that can get a raw score of 90. But if you bump that up to 50,000%, maybe only .85% of those people can get a raw score 90 on the same test. Does that make sense, a sort of diluting of the test pool with the introduction of many people that are not as interested in preparing or not fully committed to attending Law School and so just taking the test to see how it goes.

User avatar
monkeyboy
Posts: 181
Joined: Sat Dec 05, 2009 9:11 pm

Re: Most lenient curve ever?

Postby monkeyboy » Sun Feb 07, 2010 5:08 pm

twintipping_bumps wrote:So do you guys think there is that much of a lag time between experimental sections and their appearances on real tests, i.e. sections from the 40s only beginning to appear now. Some people yesterday were saying they saw sections they had recently had as experimental sections. That does seem real quick, though. If I had a RC I had had as an experimental, I don't think that would be right. I would remember alot of it.

But I guess I haven't worked out my thoughts on the curve process either. I think the performance on the experimental sections has to be the main determinant, but there must be some other formula to account for amount of test takers and the quality of performance that is statistically likely for a given number of people. I mean, if you have 30,000 people, you may get 1% of people that can get a raw score of 90. But if you bump that up to 50,000%, maybe only .85% of those people can get a raw score 90 on the same test. Does that make sense, a sort of diluting of the test pool with the introduction of many people that are not as interested in preparing or not fully committed to attending Law School and so just taking the test to see how it goes.


I think LSAC has probably been tweaking it toward a higher level of objective difficulty starting in the 49-53 range, and I would guess that they have probably been trying to stay ahead of the population that prepares heavily. I would imagine that there is strange gap between those who prepare a great deal and those who prepare very little or not at all. I think there must be almost two distributions that have to be reconciled. The top scorers among those who prepare little probably score in the 157-164 range, while the low end scorers among those who prepare a great deal probably score in the high 156-160 range. This is all conjecture, but it kind of makes sense. I am by no means a statistician and this could all be inane drivel that i'm spewing, but it seems that LSAC has two large groups of test takers to deal with in creating a predetermined distribution that makes sense. I would bet that there are strange bunchings of scorers along the raw score spectrum that weren't seen 10 years ago when a high level of preparation wasn't as common.

twintipping_bumps
Posts: 196
Joined: Tue Jun 02, 2009 11:14 am

Re: Most lenient curve ever?

Postby twintipping_bumps » Sun Feb 07, 2010 5:21 pm

It'd be an interesting book to read. I have to think it is out there: The Effects of The Test Prep Industry on the Construction of Standardized Tests and the Formulas Used to Assess Them.

User avatar
vanwinkle
Posts: 9740
Joined: Sun Dec 21, 2008 3:02 am

Re: Most lenient curve ever?

Postby vanwinkle » Sun Feb 07, 2010 5:24 pm

monkeyboy wrote:The top scorers among those who prepare little probably score in the 157-164 range, while the low end scorers among those who prepare a great deal probably score in the high 156-160 range.

Totally disagree with this. When I was doing a Princeton Review class, there were people there who were testing in the high 130s and low 140s to start, and they improved up into the low 150s (150-153 range) and were very happy to have gotten that far.

User avatar
monkeyboy
Posts: 181
Joined: Sat Dec 05, 2009 9:11 pm

Re: Most lenient curve ever?

Postby monkeyboy » Sun Feb 07, 2010 5:27 pm

vanwinkle wrote:
monkeyboy wrote:The top scorers among those who prepare little probably score in the 157-164 range, while the low end scorers among those who prepare a great deal probably score in the high 156-160 range.

Totally disagree with this. When I was doing a Princeton Review class, there were people there who were testing in the high 130s and low 140s to start, and they improved up into the low 150s (150-153 range) and were very happy to have gotten that far.


As I said, I was kind of guessing, but I am surprised that college graduates who have been tutored are scoring that low. I still think the volume of people preparing has possibly thrown a huge wrench into the curve, as the bands have probably changed. Maybe not though. Like I said, I am only speculating.

twintipping_bumps
Posts: 196
Joined: Tue Jun 02, 2009 11:14 am

Re: Most lenient curve ever?

Postby twintipping_bumps » Sun Feb 07, 2010 5:31 pm

I still wonder what the most lenient curve ever was, though. I don't have the actual 10 or next 10 books. Maybe I will stop by a book store and look at them when I have a chance.

woodstocker
Posts: 40
Joined: Sat Dec 26, 2009 12:14 am

Re: Most lenient curve ever?

Postby woodstocker » Sun Feb 07, 2010 5:47 pm

twintipping_bumps wrote:I still wonder what the most lenient curve ever was, though. I don't have the actual 10 or next 10 books. Maybe I will stop by a book store and look at them when I have a chance.


http://powerscore.com/lsat/help/correct_targeted.cfm

this might help.

User avatar
existenz
Posts: 927
Joined: Tue Dec 22, 2009 3:06 am

Re: Most lenient curve ever?

Postby existenz » Sun Feb 07, 2010 5:54 pm

woodstocker wrote:http://powerscore.com/lsat/help/correct_targeted.cfm

this might help.

Wow, test #23 had a -16 curve! On that test you could miss 4 and still get a 180!
monkeyboy wrote:
vanwinkle wrote:
monkeyboy wrote:The top scorers among those who prepare little probably score in the 157-164 range, while the low end scorers among those who prepare a great deal probably score in the high 156-160 range.

Totally disagree with this. When I was doing a Princeton Review class, there were people there who were testing in the high 130s and low 140s to start, and they improved up into the low 150s (150-153 range) and were very happy to have gotten that far.


As I said, I was kind of guessing, but I am surprised that college graduates who have been tutored are scoring that low. I still think the volume of people preparing has possibly thrown a huge wrench into the curve, as the bands have probably changed. Maybe not though. Like I said, I am only speculating.

Objectively, I think the test has become more difficult since the tests in the 30s (which, let's remember, were the early 2000s and not that long ago. Scoring a 175 on PT 40+ (and especially 50+) is harder than the tests in the 30s, even if LG are somewhat easier now (though yesterday + June 2009 being an exception to this).

I'm guessing that the LSAT tests in the 60s will be even more difficult. Imagine the LG from yesterday, the LR from December, and any recent RC all in the same test. Ouch.

LSAC knows the exact percentage of scorers, and I doubt they would not make adjustments if 5% of scorers were suddenly hitting 170+. They would either tighten the curve, which they probably don't want to do (nobody likes a -8 curve), or make the test harder. Clearly they are making the test harder.

I do have to wonder if any tweaking of the curve happens after they run all the answer sheets through their scantron machines. I know they pre-curve the test, but I wonder if there is any wiggle room if only, say, 1% scored 170+ rather than the expected 2.5%.

twintipping_bumps
Posts: 196
Joined: Tue Jun 02, 2009 11:14 am

Re: Most lenient curve ever?

Postby twintipping_bumps » Sun Feb 07, 2010 6:09 pm

Thanks for the link. Interestingly enough, the December 1998 test is bordered by 14 and 13 scales, assuming 101 questions. Granted, though, looking at this, the trend towards tighter curves is noticeable. Yet, this past December's curve seems to be the loosest since December 2004, although that was followed by a raw score of 90 for its June test.

I really believe there must be some work done to the curve after the test has been administered: at least slight adjustments to any discrepancy between actual scores and the projections from experimental sections. But most people seem to confidently assert that no such work on the curve takes place, that it is all predetermined. If I believe that, then that is why I am wondering if there is some sort of formula included in the curve calculation that accounts for total number of test takers, since I just don't see how the same percentage of people who can score say a raw total of 90 holds up as the total number of takers expands rapidly in a small window of time.

twintipping_bumps
Posts: 196
Joined: Tue Jun 02, 2009 11:14 am

Re: Most lenient curve ever?

Postby twintipping_bumps » Sun Feb 07, 2010 8:23 pm

existenz wrote:I do have to wonder if any tweaking of the curve happens after they run all the answer sheets through their scantron machines. I know they pre-curve the test, but I wonder if there is any wiggle room if only, say, 1% scored 170+ rather than the expected 2.5%.


Has anyone ever provided a bullet-proof answer on this? I've been following the board since last June, and though I have seen some people assert rather aggressively this is not the case, I haven't really been convinced myself.

I mean, in grade school, when you set a curve, you average the results and then subtract or add points to meet the desired curve. It seems as simple as that.
Last edited by twintipping_bumps on Sun Feb 07, 2010 10:03 pm, edited 1 time in total.

User avatar
vanwinkle
Posts: 9740
Joined: Sun Dec 21, 2008 3:02 am

Re: Most lenient curve ever?

Postby vanwinkle » Sun Feb 07, 2010 8:38 pm

twintipping_bumps wrote:Has anyone ever provide a bullet-proof answer on this? I've been following the board since last June, and though I have seen some people assert rather aggressively this is not the case, I haven't really been convinced myself.

I mean, in grade school, when you set a curve, you average the results and then subtract or add points to meet the desired curve. It seems as simple as that.

Yes, because 2 plus 2 equaling 4 in grade school means algebra is easy.

twintipping_bumps
Posts: 196
Joined: Tue Jun 02, 2009 11:14 am

Re: Most lenient curve ever?

Postby twintipping_bumps » Sun Feb 07, 2010 8:44 pm

Well, my thought is that curving a test like the LSAT is so complex that why wouldn't they also view the results of each administration before setting its definitive curve? It seems like to keep the results authentic from one test to the next they would need to use all information possible: results from experimental sections and how people actually performed given the specific point in time, history, and culture of the administration.

User avatar
vanwinkle
Posts: 9740
Joined: Sun Dec 21, 2008 3:02 am

Re: Most lenient curve ever?

Postby vanwinkle » Sun Feb 07, 2010 8:52 pm

Here, the information you want. From http://www.lsac.org/pdfs/InformationBookweb.pdf page 36:

LSAC wrote:Adjustments for Variation in Test Difficulty

All test forms of the LSAT reported on the same score scale are designed to measure the same abilities, but one test form may be slightly easier or more difficult than another. The scores from different test forms are made comparable through a statistical procedure known as equating. As a result of equating, a given scaled score earned on different test forms reflects the same level of ability.

"Equating" is not the same thing as curving in the grade-school sense as you described. For a good explanation of equating, there's this website which dispells myths about the LSAT: --LinkRemoved--

napla.org LSAT Myths wrote:6. Some LSAT forms are easier than others.

Each LSAT form is written to a common set of test specifications--specifications that describe both the content of questions and the distribution of questions across the spectrum of difficulty levels. Each scored LSAT question is pre-tested twice--once to gather data about how the item functions on its own, and a second time as part of an intact test section. Data from these pretests allow LSAC to 'equate' each LSAT form. Equating is a statistical process through which the very slight differences in difficulty across LSAT forms can be mitigated, thus allowing direct comparison of results from different tests. Therefore, a December 1998 LSAT score of 150 means the same thing as a 150 from the October 1995 administration, or from any administration since June 1991.

7. The LSAT is graded to a curve, so your score can be influenced by the other test takers with whom you test.

Some candidates mistakenly believe that they will be graded on the LSAT in relation to others who take the test with them. In fact, all LSAT scores are equated back to the original base form, given in June 1991. It is possible, although extremely unlikely, for everyone taking the test on the same day to earn a score of 180. Effectively, an individual test taker's performance is compared to the performance of all test takers since June 1991, except those taking the test at the same time.

Equating ensures that your score means the same thing as it would have on a prior test, not that you are exactly the X percentile against your peers in that group.




Return to “LSAT Prep and Discussion Forum”

Who is online

Users browsing this forum: Baidu [Spider], earldasquirel, Google [Bot], jackstraw92, Tomthecat, xtremenite and 9 guests