How is the curve even determined? Forum

[honestabe84]

I have yet to see a consensus on how the curve is determined. Yes, I realize it is "equated." Does this mean that it is determined ENTIRELY ahead of time or what? Can it be adjusted after the test? People seem to continuously disagree about this.

[alicrimson]

Very curious about this as well, as I hear different things from most everyone I talk to. its so shrouded in damn mystery. LSAC, fucking throw us a bone here.lol

[thesealocust]

edited

[09042014]

[Answer: Absent really strange results, a question needing to be thrown out, etc. they set the curve based on prior administrations of experimental sections]

^^^^^^^^^^

Thread over.

[3|ink]

120 - 180 converted figures are supposed to represent percentiles. Every number between 120-180 reflects a different percentage. However, the precentages are always the same for the same LSAT score. For instance, a 164 is always 90%.

Thus, the curve is calculated based on how others performed.

Edit for clarity: Specifically, the curve reflects how many people scored a raw score (out of 100 or 101) below yours.

[WestOfTheRest]

[Answer: Absent really strange results, a question needing to be thrown out, etc. they set the curve based on prior administrations of experimental sections]

This actually isn't a complete answer. Each question is rated based on it's difficulty, determined by the experimental section. The test is then equated to the difficulty of the first lsat in modern lsat terms which is a test in 1993 (I can't remember which one exactly). Then the curve is determined based on the difficulty of the current test in comparison to the original test.

PS: the curve is not determined based on how everyone performed, and lsat score percentiles have changed over the years.

[09042014]

120 - 180 converted figures are supposed to represent percentiles. Every number between 120-180 reflects a different percentage. However, the precentages are always the same for the same LSAT score. For instance, a 164 is always 90%.

Thus, the curve is calculated based on how others performed.

Edit for clarity: Specifically, the curve reflects how many people scored a raw score (out of 100 or 101) below yours.

You don't know what you are talking about. This is wrong.

[OperaSoprano]

120 - 180 converted figures are supposed to represent percentiles. Every number between 120-180 reflects a different percentage. However, the precentages are always the same for the same LSAT score. For instance, a 164 is always 90%.

Thus, the curve is calculated based on how others performed.

Edit for clarity: Specifically, the curve reflects how many people scored a raw score (out of 100 or 101) below yours.

The percentiles do fluctuate, but not all that much. A 164 was 91st percentile when I got it in October 2008.

[3|ink]

120 - 180 converted figures are supposed to represent percentiles. Every number between 120-180 reflects a different percentage. However, the precentages are always the same for the same LSAT score. For instance, a 164 is always 90%.

Thus, the curve is calculated based on how others performed.

Edit for clarity: Specifically, the curve reflects how many people scored a raw score (out of 100 or 101) below yours.

You don't know what you are talking about. This is wrong.

Please enlighten us.

My test prep company said this is how it was done. I assume they know what they're talking about, or at least more so than a typical internet 'nuh uh' guy.

[09042014]

120 - 180 converted figures are supposed to represent percentiles. Every number between 120-180 reflects a different percentage. However, the precentages are always the same for the same LSAT score. For instance, a 164 is always 90%.

Thus, the curve is calculated based on how others performed.

Edit for clarity: Specifically, the curve reflects how many people scored a raw score (out of 100 or 101) below yours.

You don't know what you are talking about. This is wrong.

Please enlighten us.

My test prep company said this is how it was done. I assume they know what they're talking about, or at least more so than a typical internet 'nuh uh' guy.

A process called equating. The guy with the roulette wheel explained it correctly.

[WestOfTheRest]

120 - 180 converted figures are supposed to represent percentiles. Every number between 120-180 reflects a different percentage. However, the precentages are always the same for the same LSAT score. For instance, a 164 is always 90%.

Thus, the curve is calculated based on how others performed.

Edit for clarity: Specifically, the curve reflects how many people scored a raw score (out of 100 or 101) below yours.

You don't know what you are talking about. This is wrong.

Please enlighten us.

My test prep company said this is how it was done. I assume they know what they're talking about, or at least more so than a typical internet 'nuh uh' guy.

How I said it is done is exactly how it is done. Save some of the more intricate statistical methods used. You want to find out about it there a plenty of papers available online about it.

[Tautology]

I just want to point out that Desert Fox is far from typical.

[eldizknee]

.

[honestabe84]

Ok, so it can or cannot be adjusted after any given administration?

[zworykin]

Ok, so it can or cannot be adjusted after any given administration?

Hard to answer. In the past they've removed individual questions from scoring after a test. It's rare, but it happens. In that case they have no choice but to adjust the "curve." If they've ever adjusted the curve without throwing away questions, based on the actual takers' performances, I don't know how any of us would ever be aware of it. My gut feeling is that they wouldn't do it unless the overall scores were well below the projected scores (like, by more than a standard deviation on average--maybe even 2 SDs).

[09042014]

Ok, so it can or cannot be adjusted after any given administration?

Hard to answer. In the past they've removed individual questions from scoring after a test. It's rare, but it happens. In that case they have no choice but to adjust the "curve." If they've ever adjusted the curve without throwing away questions, based on the actual takers' performances, I don't know how any of us would ever be aware of it. My gut feeling is that they wouldn't do it unless the overall scores were well below the projected scores (like, by more than a standard deviation on average--maybe even 2 SDs).

I'm not even sure if it take such a large score difference to make a change. I'd imagine they'd change the curve if the statistics for individual problems were significantly different from what they were predicted to be. However they are good at what they do, and I imagine it is extremely rare, and may not have ever happened.

[honestabe84]

This is why I think the creators really screwed up the curve on this test:

There was one game (the last game) that had a really ambiguous term, and everyone is complaining about it. I guarantee that when this game appeared as an experimental, it did not have that confusing term. As a result, not many (no one?) had to stop and think about a single rule and more people would have gotten the game right, or at least more questions right. So the curve for this test is not going to factor in ALL the trouble people had with that game.

Unless the creators intentionally made this game ambiguous, which I would find very hard to believe.

[Jrugg88]

It wasn't ambiguous. Reading all of the rules holistically made it obvious.

[09042014]

This is why I think the creators really screwed up the curve on this test:

There was one game (the last game) that had a really ambiguous term, and everyone is complaining about it. I guarantee that when this game appeared as an experimental, it did not have that confusing term. As a result, not many (no one?) had to stop and think about a single rule and more people would have gotten the game right, or at least more questions right. So the curve for this test is not going to factor in ALL the trouble people had with that game.

Unless the creators intentionally made this game ambiguous, which I would find very hard to believe.

How the fuck can you guarantee that? I'd bet it was exactly the same. And exactly as many people got it wrong in experimental. That doesn't mean I don't believe it was ambiguous, it might be. However every test there is a question people misread, and bitch about. And then the test comes out and everyone realizes they read it wrong.

[3|ink]

This is why I think the creators really screwed up the curve on this test:

There was one game (the last game) that had a really ambiguous term, and everyone is complaining about it. I guarantee that when this game appeared as an experimental, it did not have that confusing term. As a result, not many (no one?) had to stop and think about a single rule and more people would have gotten the game right, or at least more questions right. So the curve for this test is not going to factor in ALL the trouble people had with that game.

Unless the creators intentionally made this game ambiguous, which I would find very hard to believe.

How the fuck can you guarantee that? I'd bet it was exactly the same. And exactly as many people got it wrong in experimental. That doesn't mean I don't believe it was ambiguous, it might be. However every test there is a question people misread, and bitch about. And then the test comes out and everyone realizes they read it wrong.

I can disprove that. I took this section as experimental in December. Everything was the same. For some reason, I was able to get through most of it in December. I couldn't get past question one yesterday.

[09042014]

This is why I think the creators really screwed up the curve on this test:

There was one game (the last game) that had a really ambiguous term, and everyone is complaining about it. I guarantee that when this game appeared as an experimental, it did not have that confusing term. As a result, not many (no one?) had to stop and think about a single rule and more people would have gotten the game right, or at least more questions right. So the curve for this test is not going to factor in ALL the trouble people had with that game.

Unless the creators intentionally made this game ambiguous, which I would find very hard to believe.

How the fuck can you guarantee that? I'd bet it was exactly the same. And exactly as many people got it wrong in experimental. That doesn't mean I don't believe it was ambiguous, it might be. However every test there is a question people misread, and bitch about. And then the test comes out and everyone realizes they read it wrong.

I can disprove that. I took this section as experimental in December. Everything was the same. For some reason, I was able to get through most of it in December. I couldn't get past question one yesterday.

That isn't proof, and I'm shocked they reused it so fast.

[3|ink]

This is why I think the creators really screwed up the curve on this test:

There was one game (the last game) that had a really ambiguous term, and everyone is complaining about it. I guarantee that when this game appeared as an experimental, it did not have that confusing term. As a result, not many (no one?) had to stop and think about a single rule and more people would have gotten the game right, or at least more questions right. So the curve for this test is not going to factor in ALL the trouble people had with that game.

Unless the creators intentionally made this game ambiguous, which I would find very hard to believe.

How the fuck can you guarantee that? I'd bet it was exactly the same. And exactly as many people got it wrong in experimental. That doesn't mean I don't believe it was ambiguous, it might be. However every test there is a question people misread, and bitch about. And then the test comes out and everyone realizes they read it wrong.

I can disprove that. I took this section as experimental in December. Everything was the same. For some reason, I was able to get through most of it in December. I couldn't get past question one yesterday.

That isn't proof, and I'm shocked they reused it so fast.

It isn't definitive proof, but it is evidence (if we're going to be sticklers). The 3rd game was in my December experimental as well.

[09042014]

It isn't definitive proof, but it is evidence (if we're going to be sticklers). The 3rd game was in my December experimental as well.

Well, you aren't claiming to remember the wording was different, just that you did much worse on it right? It's very likely you just flubbed it this time around.

[Jessss22]

120 - 180 converted figures are supposed to represent percentiles. Every number between 120-180 reflects a different percentage. However, the precentages are always the same for the same LSAT score. For instance, a 164 is always 90%.

Thus, the curve is calculated based on how others performed.

Edit for clarity: Specifically, the curve reflects how many people scored a raw score (out of 100 or 101) below yours.

You don't know what you are talking about. This is wrong.

this is completely wrong. 75 question right on one exam and 75 questions right on another exam can produce totally different score

[whatjusthappened]

This is why I think the creators really screwed up the curve on this test:

There was one game (the last game) that had a really ambiguous term, and everyone is complaining about it. I guarantee that when this game appeared as an experimental, it did not have that confusing term. As a result, not many (no one?) had to stop and think about a single rule and more people would have gotten the game right, or at least more questions right. So the curve for this test is not going to factor in ALL the trouble people had with that game.

Unless the creators intentionally made this game ambiguous, which I would find very hard to believe.

How the fuck can you guarantee that? I'd bet it was exactly the same. And exactly as many people got it wrong in experimental. That doesn't mean I don't believe it was ambiguous, it might be. However every test there is a question people misread, and bitch about. And then the test comes out and everyone realizes they read it wrong.

I thought that they change the game up a lot when it goes from being an experimental to the real thing (i.e., different scenario)?