I'll just comment on our tests (ManhattanGMAT).

- Our tests have a standard deviation of 50 points, compared to a standard deviation of 30 points on the official test. So, close, but not as good as the real thing (obviously it would be impossible to be the same as or better than the real thing!).

- Our algorithm mimics the algorithm used by the official test, though we do not have experimental question capability built in yet (though we expect to be the first prep company to have this, hopefully next year sometime).

- We run statistical tests of every single question to determine the question's validity, predictability, difficulty level, and about 5 other parameters. Any that fail even one of those tests gets pulled from the database and either fixed or discarded. As far as I'm aware, no other company continually checks and tweaks its question pool in this way (and between me and my colleagues, we have worked in the past for all of the major and many of the minor competitors out there). In fact, this can only be done for tests that are given online, so that the database can be changed dynamically. Tests taken from CDs obviously can't do this.

- We also ask our students to report their official results to us and we run backward tests to check the accuracy of their scores on our tests vs. what they got on the official test. We use that to tweak our scoring algorithm; this is what allows us to get our standard deviation so close to that of the official test.

I have had students in the past with Macs and they haven't told me that they can't access the tests, but I would check with our student services team to make sure, if I were you. [email protected]

Also, we don't have standard discounts or anything (just buying access to the tests is really not that expensive - I think it's like $40 or $50) but we are running a promotion right now. If you attend one of our free Open House info sessions, you get free access to the tests for one month. We hold the sessions in-person in several major cities around the US and we also hold them online via our web conferencing platform, so you can attend from anywhere in the world (you do need a broadband connection, though).

I'm not sure how long the free promotion will last, so if you're interested, go to our web site and look for the "free events" link to sign up right away.

I know TPR's tests are technically adaptive but I'm not sure how much they mimic the actual test. I don't know how closely they follow IRT (the algorithmic theory behind the official test). Oddly, they haven't just offered to share their algorithm with me... of course, I haven't asked... lol.

Anyway, from some stats that people have posted on here before, though... well, IMO some things seem kind of off as far as the theory is concerned. But I don't know for sure as I haven't had an opportunity to examine their actual algorithm.