<p>There seems to be disagreement here on CC about this. Do they look at the test and determine ahead of time what the curve will be, or do they determine it after receiving scores? I would think the latter, for a few reasons. One, if you were the collegeboard and already knew that a section of the test is abnormally hard or easy before administering it, wouldn't you adjust up/down it so it was perfect, negating the need for curving? Two, the wording on their website seems to indicate the post-test curving technique. Three, it seems more logical. After they receive the scores and look at it on a normal distribution (bell shaped curve) and compare it to past tests, they can see if the distribution is skewed in an unnatural way to the left or right. This is a simplified version of the statistical analysis they probably preform, but nonetheless makes my point. What do all of you think?</p>
<p>The College Board uses information from the equating section on a specific SAT test date to determine the score scale (out of 800) after the test is administered.</p>
<p>The SAT currently employs two forms of equating: nonequivalent groups anchor test design and random/equivalent groups design. Put simply, it’s a method by which the difficulty level of the specific form of the SAT on a given test date is “normalized” to previous versions of the test. This process is performed to ensure that, on any given test date, a student does not have a scoring advantage (versus previous test dates).</p>
<p>Conduct a Google search for a PDF entitled “Equating Test Scores” by Samuel A. Livingston. Read the document. It’s well-written and explains the nuts and bolts of the equating process.</p>
<p>Equating test scores has to do with making sure that scores on one form of a test conform with scores from other forms of the test, including those taken on different test dates, even though one test has different questions from another. College Board usually does that after a test is taken even though for many tests it could be estimated beforehand because CB often gives a test that has been given at times in the past. A different issue is to whom you are actually being compared to determine your percentile ranking. As College Board has also said, to determine a student’s percentile rank for any test taken during an administrative year (Oct to June), the student’s scores are compared to the scores of the previous year’s graduating seniors, see “Percentiles” on this page: [Understanding</a> Your SAT Scores](<a href=“The SAT – SAT Suite | College Board”>Understanding SAT Scores – SAT Suite | College Board).</p>
<p>Thank you, that clears things up. I’ve always understood why the college board equated and more or less how (and have agreed/disagreed with the policy at varying times in my career), but the scaling afterwards is the most logical. Also explains why it takes so long to get scores back.</p>
<p>
</p>
<p>Exactly (since the questions are all pretested – experimental section – and CB knows with near certainty which questions are easy, medium and hard. Thus, they can predict, with near certainty, how many of which questions will be missed on any given test date.</p>
<p>An easy way to think about it is using Sat-Math as an example; SAT-M includes 6-8 Alg II questions. Assuming all Alg II questions are of the difficult category, a test with 8 such questions would be ‘harder’ than a test with only 6 Alg II questions. Thus, by equating, one can miss more of the harder test and still score high.</p>
<p>Bluebayou, it makes sense that they use experimental questions that pass the statistical analysis on later tests, some people who took the March and June tests saw their experimental as a real passage nearly verbatim. But what doesn’t make sense is if they know a test is unusually hard or easy (ex. 10 hard algebra 2 questions in a math section) before administering the test, why don’t they alter it before hand to make it closer to a normal test?</p>
<p>
@neuromajor:</p>
<h1>1: If your friends saw the same experimental section on the March and June tests…they should have tanked the section on the June test. Think about it. The “harder” the experimental section is…the more favorable the scaling, right? For better or worse, a small group of smart kids probably wouldn’t move the scaling in any significant way. :-)</h1>
<h1>2: The example that bluebayou gave was hypothetical (and so was yours). I suspect that there is very little variation in difficulty level from test to test…and that is fairly consistent with the published score scales (Question-and-Answer Service) for the last few years.</h1>