Accuracu of MGMAT's scorring algorithim

[daniel.ndegwa]

I took a GMATPrep and an MGMAT CAT test over the weekend and scored the same in both i.e. 38Q and 35V -610.
What I dont seem to reconcile is my level of accuracy in the GMATPrep test (Verbal was 30 question correct and Quants was 22 questions correct. Compare this to MGMAT CAT where verbal had 24 questions correct and Quant was 18 questions correct.
Not sure how MGMAT works because of this discrepancy but would be interested to know its predictive ability for the real GMAT score.

Many thanks

[StaceyKoprince]

It isn't the case that the test is scored primarily based upon number correct - that's actually only a marginal consideration. You can have different numbers correct and still get the same score, depending upon the distribution of right and wrong answers, the difficulty levels of the right and wrong answers, and so on. You can also have the same number correct and get a very different score, again depending upon many factors.

Basically, that old thing we learned from school about "get more right and your score goes up" - that doesn't apply to computer adaptive tests. :)

The standard deviation on our test (between tests) is about 50 points (compared to about 30 points on the real test). The standard deviation from final practice test to official test for our students is also about 50 points (note that this second figure includes only those students who report their official test scores to us).

The standard "skew" from final practice test to official test across all students (who report scores) is less than 10 points. That is, our scores do not skew consistently higher or lower across all students, though individual students will obviously see greater variation.