The early questions to worth more right....?

I am neither denying the rewards of GMAT nor doubting the abilities of the fellow test takers. So many elements are capricious with this test and for many doubts "Because I said so" seems to be the answer.

I really want to understand how can one be assigned a 90th percentile when the fellow 99 test takers do not take the same set of questions. Lets say I am being an idiot for a second, Look at LSAT and MCAT which are still PBT and CBT. Do you think AAMC and LSAC does not have the $$ or infrastructure to roll out CAT? Why havent they done so after 5-6 years of GMAT's CAT and why dont we even hear a word about it. It sounds pretty ludacris to me to say that 200,000 folks took similar questions on a test and hence we can arrive at the conclusion that these percentiles are valid. GMAT might be using the test taker's responses to calculate the difficulty of a question, but those responses might have come under time crunch and possibly the test taker's familiarity with particular topic. Its not perfect is all I am trying to say and there is enough capriciousness in this test.

Hades, Do not take any offence. I am not talking about your skills or abilities here. I have seen contradicting stuff from GMAT and honestly believe that it has some issues. Thats what I am talking about.

As you said lets say, You got all the difficult ones correct. that means most of your mistakes came from easy/medium questions. Then why did the test give difficult questions to start with when a test taker gets easy/medium ones wrong?

I got 48 with 11 wrong and 49 with 10 wrong. In fact, I got 5 of the last 7 seven wrong by answering fluke and still got 48. I dont know whether that will hold on the actual test or not as it did on the GMATprep.

Generally speaking, high scorers get more difficult questions than do low scorers. However, even high scorers get easy and medium questions. Try going GMATPrep and getting all the questions wrong- you will still see some difficult questions. The reverse is also true.

What's contradictory about 'The first 10 questions do not count more than the rest' from the horses mouth?

Write a GMATPrep, get all the easy-medium questions wrong but get all the difficult questions correct, now this is hard to do as you have to be able to discern the level of difficulty from a question, but it can be done

Also another note, if you spend say 25 minutes on the first 10 and get them all correct, and have 50 minutes for the last 27 and due to the lack of time end up getting quite a few wrong, that will penalize you a lot more than say doing the first 10 in 15-20 and having more time for the last few

But again don't try to predict the algorithm... there is no substitute for learning the material inside out and mastering the difficult level questions.

I definitely agree with the last sentence. If I were not trying to learn, I would not be here.

Keeping learning aside, My point is that the test has to be fair in all respects. It is my opinion, you might disagree, that the test can be capricious and unfair to an extent that it affects scores.

Because every test taker takes the exam with different set of questions, can GMAT assure that the same test taker taking the test two times (with in a week and with out much preparation and assuming that other conditions are identical) the same score or with a 10 point difference ? If that's not the case the test is not 100% accurate and the whole percentile thing does not make sense.

I am not the first one to talk about this and will not be the last either. Just because I am trying to understand how things work does not mean that I am making predictions my priority. The reason why I want to know is this is because, a score of 47 and 48 on Q or as one reaches higher scores in Q/V, every single decision the test takes will affect your score. GRE does let you see the questions you got wrong. I have raised so many points and no one seems to respond to them. All that I hear is because "GMAT said so. Now go back to your book"

Do you guys think that schools such as Stanford, Sloan are accepting GRE just for fun? While attracting students in remote places is definitely a key goal in accepting GRE, there are issues bigger than that. Just wait and see in the next few years how many schools will join the parade behind stanford/sloan/Harvard. I believe 2013 will be a tad bit late for GMAC to get to the new pattern and start testing some thing else.

They accept GRE in addition to GMAT for a number of reasons (note that they of course still do take the GMAT)

1. It can be written in certain locations where the GMAT is unavailable, ie foreign countries
2. To get more applications (read more $$ & more rejections, hence higher selectivity and an advantage in the rankings, which leads to more applications)


And it has nothing to do with the GRE being a better approximator than the GMAT

But at the end of the day if you score 700 on the GMAT you definitely have some intelligence, and that's what bschools value (along with personality/$$ making abilities).

'GMAT says so'

That's how it is, get used to it

I think the %ile rankings are generally pretty reflective of a person's overall skill level. You mentioned LSAT & MCAT before and how they aren't electronically administered. Well how is a 2006 170 LSAT the same as a 2009 170 LSAT? There is a large enough body of questions that a paper-test system isn't perfect either.

How do people score usually within just 20 points of their GMAT Prep practice scores? They never see the same questions on the real thing. But their scores don't really deviate. The reason lies in the accuracy of the computer to determine your scores. When GMAC is testing literally thousands of questions and they have millions of responses from hundreds of thousands of test takers, they can create a very high confidence interval that a question is a particular difficulty.

You made a very good point that 170 on the LSAT in 2006 is not the same as 170 on the LSAT in 2009. I agree, But The tests and the results for LSAT are only for that test. So all those test takers answer the same questions and those who took the test in Feb are taken into percentile allocation. Right? If LSAT assigns percentile based on the whole 4 administrations in a single year or over 2/3 years, then I am up for some enlightenment. But Hey If I am wrong, I am wrong. Yeah! PBT has its own challenges.

I agree that GMAT might have high confidence interval but it might just not be accurate for every test taker for every single question. I feel that it makes a difference after the 80th percentile scores.

There was a study done a few years by the gmac that compared GRE with GMAT and the standard error in score correlation was 110 points.
That tells me that either gmat and gre don't correlate well or that the test mechanics are not perfect. You can take either side.

Posted from my mobile device

The verbal section of the GRE is quite different. There is much more emphasis on vocabulary, though attempts have been made to change this.

I beg to differ. I'm in no way questioning your abilities. This is just my perspective.

We'll know about experimental questions, both in the Quant & the Verbal section - some believe, others don't. I do.

For experimental questions, GMAT does not differentiate between easy or difficult questions. It just throws it out there. So, answering all difficult questions does not guarantee that you aced the test and you're bound to get a high score. You scored Q50, may be, because you got all the *counted* questions correct.

I've scored Q48 with 13 wrong - errors in Probability, Combination and Co-ordinate Geometry (PS) and Inequalities & RTD questions (DS), in the same test. I'm assuming we consider most of these topics as difficulty.

I had a great opportunity to speak with a psychometrician from GMAC at a conference. He went into great detail about how the CAT calculates the score. It is actually a fairly in depth series of calculations. It would require quite a few diagrams (I used to have some of them on disk, but I have to see if I can find them).

So the short answer is that technically, yes, the earlier questions are "worth more," but ultimately that is a very oversimplification of the process. Ultimately, people who try to "game the test" shouldn't be able to succeed at it, so worrying too much about this idea could ultimately hurt you more than help you. As another poster pointed out, if you are so worried about the early questions that you run out of time in the end, your score may actually be hurt more by that performance than had you worried about doing your best on the questions you see while ultimately being able to spend time on all the questions.

The psychometrician also did actually tell me that the algorithm used on the GMATprep software is exactly the same as the one used on the real exam. In fact, he said that he hand picked the questions for the GMATprep to utilize some of the most discerning questions for evaluating test scores.

I will try to find some of the data he provided at the conference and post it later. Otherwise, I will just write up a lengthy post about how questions are assessed and how that affects your score in more detail.

Good to know Let us know what you dig out or post some thing more you know.

I would also be interested in seeing this diagram if you can find it.

There is also a relevant thread that was started by BSD_Lover last year titled (Un) Scientifid Analysis of GMAT Prep
un-scientific-analysis-of-gmatprep-64970.html