standard deviation of MGMAT and real GMAT exams

[tomslawsky]

I have read from this board that the SD of the MGMAT is 50 and the real GMAT is 30. Is this the correct interpretation of this data:

Real GMAT score 700 means you have a 67% chance of scoring 670-730 (+- 1SD), a range of 60 points

MGMAT score of 700 means you have a 67% chance of scoring 650-750 (+-1SD), a range of 100 points?

If this is the correct interpretation, is there a statistical confidence interval associated with these numbers?

If this is not the correct interpretation, can you please correct me. Thank you.

[StaceyKoprince]

Yes, this is the correct interpretation, but I want to provide more detail. There are two different things being measured here.

The SD published for the real test is what's called a "within-student" SD, meaning that if that student were somehow able to take 100 tests in a row, today, then that student could expect 2/3 of the scores to be within +/- (about) 30 points of the given score and 1/3 of the scores to be outside of that range.

The equivalent "within-student" SD on our tests is about 50 points.

We have also calculated the standard deviation between our students' final practice test and the score they report to us for the real test. That figure also happens to be about 50 points.

I don't know what the confidence interval is - good question! I'll ask.

[jeffboaz]

Stacey

I'm curious..... comparing students MGMAT final test score to their actual GMAT. What percentage of students score higher/lower on the real test?

Thx
Jeff

[StaceyKoprince]

It's about 50/50. We normalize our scoring algorithm every two months to ensure that the 50/50 mark is within +/- 5 points of even, so that there is not a general bias toward scoring either higher or lower, on average, on the real test.