gcf101

^What's cool about this monkey-wrenching scenario is that if one takes a test with that insouciant crowd his percentile will hit 99% even with the score of 1200.
BigIs and xiggi must be sick and tired of debunking the "best test date" theory in umpteenth time.
Here's a good old post:
Question about Test date

QuantMech

. . . at least for some. I skimmed the report, linked by fignewton. The "correct" operation of the equating procedure (to ensure that the same level of performance yields the same score, on different test dates) is based on the assumption: For the students testing on a given date, the performance on the equating section is accurately representative on their "true" performance on the test as a whole.

We know that the equating sections must contain questions that have been used multiple times. This will tend to create some level of familiarity with these questions in the test-prep industry. Therefore, I'd hypothesize that students who have been extensively prepped might actually perform better on these questions than they do on the test as a whole (since presumably the whole test will contain some questions of a less familiar type). Because of the equating procedure, this will tend to make the testing population look "smarter" than they really are, thus shifting the mean up and giving people a small boost.

So, suppose that you are generally talented at test-taking, but you have not prepped. If my hypothesis is correct, I believe you should aim for dates when the largest numbers of highly prepped students are taking the test. Here's my thinking: The highly prepped students will perform better relative to you on the equating section, but they will not perform as well on the remainder of the test--by this, I do NOT mean that they will outperform you on the equating section, just that the ratio of their score to yours is likely to be higher on that section than on the rest. Your "underperformance" on the equating section does not matter to your score. Meanwhile, the "overperformance" on the equating section by the prepped students will boost the score on the 200-800 scale that is assigned to the mean, and this will tend to yield a boost for you, wherever you fall on the scale. In addition, the pattern of performance by the prepped students might actually reduce the apparent standard deviation within the testing population, thus giving you an extra boost for scoring near the top.

If CB could obtain accurate data on the level of test-prep for individual testers, they could determine whether or not my hypothesis is correct.

QuantMech

^^Last sentence, first paragraph, should be "accurately representative of . . . "
(The computer I'm using now is fire-walled to the teeth, and I don't have the "Edit" option available.)

fignewton

I don't really see how someone could attain a "better level of familiarity" with questions from the equating section, whether extensively prepped or not. The questions are never released (except when being retired). Conceivably you could raise your familiarity by taking the SAT many times, but the CB surely has a large pile of equating section questions, and you can be sure that the equating questions that show up on a given test haven't been used for a while. You would have to cook up a scenario like: someone with a photographic memory would have to write down all the questions on her test and give them to a brother or sister for use in a couple of years.

tjan91

March 2008 - easiest

QuantMech

It's a hypothesis, potentially testable by CB, though not by me.

But I'd surmise that the people writing test-prep questions for the prep companies have gained some familiarity with actual SAT questions by taking the test multiple times. Simply quoting the questions would be a copyright violation (I believe); but setting up questions that are similar in feel, or that have similar tricky elements is presumably permissible. For example, I think that in certain specific contexts, test-takers often overlook the possibility that a variable, x, might be negative; so the prep companies alert their students to that via their practice questions. Then their students would have probably have better-than-typical odds on related questions, on the real SAT. Also, I'm continuing to advance the hypothesis that the advantage winds up being slightly greater on the equating section than on the test as a whole. (For example, I'm not sure that Algebra II is featured in the equating section yet.)

Also, there have been reports on this forum of tests being repeated from a Sunday test date to a later Saturday test date. This suggests to me that the CB's stock of questions is not so huge. Similarly, posts in this forum have reported rumors of overseas students being "assigned" to memorize a small number of specific questions on the exams.

Wild hunch--we know that CB dropped the analogies and the quantitative comparison questions. Is it possible that there were not enough of them to ensure the integrity of the tests--because it became possible for people with good prep to anticipate some of them?

I took the SAT 38 years ago, but I can still recall a few of the questions. Also, I believe that some of the questions on a third-grade standardized achievement test my daughter took (quite a few years ago now) were substantively identical to the questions on a third-grade achievement test I took, many, many years ago. I would not have guessed this in advance!

QuantMech

An example of a geometric property that is often overlooked, where a prep course could give a student better odds of success with related questions: Circles can be co-tangent with one inside the other, as well as being externally co-tangent.

sydneysider009

I took the June test this year and it seemed a lot easier than the January test. Whether or not the scales will normalize the two testings to give me the same score, a test that seems easier does inspire more confidence and gave me more time at the end of each section to look over my answers. I don't think anyone has considered this option.

Also, I would agree with QuantMech that more practice of SAT sections does benefit people who take prep courses. They are familiar with the format, time pressure, and there are definitely questions and concepts that repeat themselves.

Finally, I have always been under the assumption that each test date is normalized with one standard deviation being the equivalent of 100 points, which would explain the variability in raw scores to scores out of 800, but would also mean that scores are based upon how smart everyone else taking the test is.