The official answer to How the CAT works??

Guyz couple of days ago i bothered you with my doubts and decided to make it clear for the others,who can ask same questions later on.

Here it is.

--------------------------------------------------------------------------------------------------------------------------------------

Official answer by Andrew Tayo on Knewton support.

Hey everyone.

Wanted to chime in.

We're all used to tests being graded by the number of questions we answer correctly. This is the kind of scoring we've experienced all through school. Even a test like the LSAT is scored by counting the number of operational questions you answered correctly, and then applying a function to transform that number-correct score to the reported scaled score - in the case of LSAT, from 120 to 180. And on a test like the LSAT, every question contributes the same amount to your overall score. But a CAT like the GMAT works very differently. While missing a lot of questions will of course affect your score, there are a couple of additional factors to keep in mind.


First, all tests like the LSAT and GMAT present some number of questions to you that are not part of the operational test. These are often called "pretest" or "experimental" questions. On the LSAT, the pretest questions are all in separate sections from the operational questions. On the GMAT, they are interspersed among the operational questions. Your performance on pretest questions does not affect your overall score. These questions are on the test so that the test-maker can gather data to determine whether the questions are good enough to use operationally. So sometimes you might miss several questions on the GMAT that don't actually count. In those cases, it looks like you missed a lot of questions, but your score is still pretty high. (By the way, there's no point in trying to guess which are the pretest questions - just try your best on every question.)

Second, it is important to remember that the purpose of the test is to sort people along the score scale. The test is designed to try to determine what score is most likely to be the one that most closely matches your current skills and abilities. It makes use of two properties of test questions in order to do that. One property is how hard or easy a question is. We all know about that one. But the other is something most people don't know, called "discrimination". Not all questions are created equal in their ability to sort people. Think of test questions as fences you have to climb. Some are low (adequate sorters), some are of medium height (good sorters), and some are really high (great sorters). Some of the fences are wooden split rails (these are the easy questions), others chain-link (medium difficulty) and still others are razor-wire (these are the really hard questions). Your score is determined not just by what kind of fences you climb, but also by how high the fences are. Roughly speaking, the higher the fence, the more accurate the sorting.

With me so far? People often think that it is more important to get the hard questions right than the easy questions, but it's not quite so simple. Imagine you are making your way through the test. After each question the CAT engine is estimating your score, based on (a) how you answered, (b) what kind of fence it was, and (c) how high it was. Say it thinks your score should be way up there, based on your performance so far. Now you come to a wooden fence that's really high, and for some reason you just get it wrong. The CAT says "Hey, my last estimate was a pretty high score. This person really should have answered this question correctly, but they didn't." BAM! it lowers its estimate of your ability, a lot more than it would have if you had failed to climb a low razor-wire fence. The higher the fence, the more it is likely to affect your score. Miss a few high wooden ones that you really should have gotten over, and your score will definitely suffer, particularly toward the end of the test, when there is little time left to pull it back up again.

Getting over high razor-wire fences causes the CAT's estimate of your ability to go up a lot. Missing high wooden fences causes it to go down a lot. All other things being equal, it does more damage to your score to miss an easy question than a hard one, and it helps your score more to get a hard question right than an easy one.

Like with pretest items, however, there's no point in trying to figure out which questions are hard or easy (fence type), and which questions sort especially well (fence height). It is really almost impossible to tell. If it could be done accurately and reliably, test-makers wouldn't need to pretest questions - they could just tell when they wrote them how hard they'd be, and save themselves a lot of money. But they can't, even though they spend all day every day writing and reviewing test questions. So don't waste your time on it - focus instead on determining what the question is really asking you to do, and then do it.

So, to reiterate: on a CAT like the GMAT, because of the way CATs work, and because of the fact that the test includes pretest questions that don't contribute anything, the raw count of questions you answer correctly/incorrectly is not what determines your score, but rather the characteristics of the specific operational questions you faced and how you answered them.