a quick question about score variation in each section Forum

[swtlilsoni]

For each section on the test, people don't usually get the same exact score every time, there is usually a variation (for example people will get 0-3 wrong on a section).
But when people have a consistent variation of +/- 1 or 2 on EACH section, that should translate to a variation of +/- 8 to 12 on the entire test! (because that variation is multiplied by 4 sections).
But it is usually uncommon for people to have that much of a variation of scores on the entire exam. Usually people are +/- around 3 on the entire exam.

For example:
it is common for people to get individual scores like this:
0-3 LR
0-3 LG
0-3 LR
0-3 RC
That variation is common.

But a person who scores that way could either get 0 wrong (if they happen to score on the high side of the variation for all sections) or 12 wrong (if they happen to score on the low side of all sections) ..
But it is EXTREMELY uncommon for people to say "Oh I consistently get between a 180 and 169 on PTs"

Usually people's full test range is considerably LOWER than their individual section ranges added up. And that makes no sense?

If it is common to get 0-3 or 0-2 variation on individual sections, why is it uncommon to get 0-12 or 0-8 variation on the entire test?
Why doesn't the variation from individual sections translate to variation on the entire exam?

[derangeddirector]

Because it is unlikely that you will do your best on every section or your worst on every section. Using your example of 0-3 misses on every section wouldn't make the test takers score swing from 180 to 169. Instead, on one test he might miss 3 on each LR, 0 on RC, and 1 on LG. That would mean he missed 7 altogether. On the next test, he might miss 1 total on LR, 3 on RC, and 3 on games. That would again mean he missed 7 altogether.

Both examples would fall within the ranges you gave even though he missed different amounts in each section. And he would have scored about the same depending on the curve for that test.

[bjlee85]

This would be assuming you have an equal chance of getting 0, -1, -2, or -3 on each individual section. This is unlikely the case; it's much more likely, at least for me, to make either -2 or -3 mistakes on a LR section, for example, although I have done a handful with 0 mistakes, and also some with -6.

God, I'm picturing this as an LR question in my head now. "Fails to consider the possibility that individual scores within a range are each equally likely to occur."

Freakin' out about tomorrow. Best of luck to everyone who's going to be taking it with me. Let's all do our best.

[pupshaw]

This would be assuming you have an equal chance of getting 0, -1, -2, or -3 on each individual section. This is unlikely the case; it's much more likely, at least for me, to make either -2 or -3 mistakes on a LR section, for example, although I have done a handful with 0 mistakes, and also some with -6.

God, I'm picturing this as an LR question in my head now. "Fails to consider the possibility that individual scores within a range are each equally likely to occur."

Freakin' out about tomorrow. Best of luck to everyone who's going to be taking it with me. Let's all do our best.

This isn't even the issue. Even if you were equally likely to get 0, -1, -2, or -3, going overall -0 or -12 would be far less likely than going, say, -6, because there are many more possible results that yield -6.

Picture rolling two dice--there is only one way to roll a 2 or a 12, but there are six ways to roll a 7. So you're going to see a lot of 7s and relatively few 2s or 12s. Same principle here.

Another way to think about this is that if we assume you are equally likely to get 0, -1, -2, or -3 in a section, then your chances of getting a -0 or -12 overall are only 1/256.

[Jeffort]

The psychometrics involved in developing and assembling each LSAT test-form are very complex. The test is designed to make sure that each achieved scaled score represents a particular ability/performance level with all the skills, concepts, etc. the LSAT is designed to measure within +/- 3 points so that test scores are statistically significant and comparable to the performance of other students achieved test scores no matter which test-form the student was administered under pretty much the same proctored conditions.

That means that no matter how the difficulty levels of the questions is balanced in the four scored sections, the same scaled score represents the same performance/ability level reliably for comparison purposes across test forms spanning the previous three years of administered tests.

You should not concern yourself with this unless you are in graduate school studying to be a psychometrician, are one or are studying complex math and statistics.

Here is an article with a description of some of the psychometrics involved in developing and assembling each LSAT test, but you should not read it or try to make sense of it if you are taking the test tomorrow or sometime in the near future unless you have time to kill and are just curious and want to get your brain boggled far more than any LSAT ever will.

http://www.math.washington.edu/~billey/ ... Week.3.pdf

BTW, the LSAT is not graded on a curve, I wish people would stop using that term. It is a test graded on an equated SCALE so that each scaled score means the same thing no matter which administered test one achieved it on within the previous three years, meaning that a 150 etc. always represents the same performance/skills/achievement level no matter which administration as long as it was taken in standard test day conditions (no extra time/special accommodations).

[derangeddirector]

BTW, the LSAT is not graded on a curve, I wish people would stop using that term.

Yeah, I realize the test isn't graded on a curve. I was simply using that term because that's what's common to indicate that a -7 might be a different score on two different tests.

Out of respect to you, I will now only use the word scaled.