GMAT Strategy

The standard deviation on our tests is about 50 points, which means your "true" scoring level is within a 50 point range on either side of the scores that you're getting. The standard deviation on the official test is about 30 points. GMATPrep hasn't published an official standard deviation, but it is either the same as the SD for the official test or a little bit higher (depending upon how closely GMATPrep really does mimic the real thing). I usually think of GMATPrep as a 40 point SD, just to be conservative.

So our tests are not quite as precise as the real thing, and probably not quite as precise as GMATPrep, but they're fairly close. In terms of practice tests, GMATPrep is the most precise one available.

The more important thing to notice is the general upward trajectory - it means you really are improving. Keep up the good work!

Stacey Koprince wrote:The standard deviation on our tests is about 50 points, which means your "true" scoring level is within a 50 point range on either side of the scores that you're getting.

I asked this in another thread, with no response: the standard deviation of what? How are you measuring test-takers' 'true scoring level'? If you are comparing MGMAT test scores with true test scores, how do you account for the improvement people will make by doing simulated tests before their real test? And what account is made for the standard error of both test scores- the simulation test, and the real test?

I'll ask our algorithm guy to give me an answer to your question and then I'll post it here. I know he compares peoples' MGMAT practice test scores with the official test scores they report to us after they've finished the course (or tutoring). Beyond that, he'll really have to get into the details of how the calculation is made.

Stacey Koprince wrote:I'll ask our algorithm guy to give me an answer to your question and then I'll post it here. I know he compares peoples' MGMAT practice test scores with the official test scores they report to us after they've finished the course (or tutoring). Beyond that, he'll really have to get into the details of how the calculation is made.

Thank you, Stacey- I ask out of genuine curiosity. It does seem a difficult thing to measure. If practice materials are genuinely useful, people will improve by using them. This means you would not expect (nor want, as a test prep company) exact correspondence between practice test scores and true test scores, at least if, as I assume is done, the practice tests are completed before the true tests. Indeed, one should only find an exact correspondence if the practice test scores are biased, or if the practice tests are pedagogically useless (which is not something I mean to suggest- it's simply a logical consequence of the analysis). So I wonder how you might account for the pedagogical value of simulation tests in the statistical analysis of your test results.

Here's Tate's response:

"The 50 pts is actually a standard error, not a standard deviation. It’s an estimate of the tracking error of the LATEST practice test prior to the GMAT, vs. the GMAT itself. That means that about 2/3 of the time, our test will predict the GMAT score within 50 pts. Note that the GMAT itself has an internal std error of about 30 points."

See, I told you he should respond, not me. I've been calling it standard deviation all this time. lol

Thanks for the response- I was interested to know the methodology behind that calculation.

Stacey Koprince wrote:See, I told you he should respond, not me. I've been calling it standard deviation all this time. lol

I understood what you meant . Standard error is just the (estimated) standard deviation of the error in a measurement, so you weren't really wrong anyway.