Daniel Doubrovkine bio photo

Daniel Doubrovkine

aka dB., @awscloud, former CTO @artsy, +@vestris, NYC

Email Twitter LinkedIn Github Strava
Creative Commons License

My son did very well on the NYC DOE Gifted and Talented test, especially considering that we did no prep other than exercising with the DOE samples from the handbook, maybe a couple of times to teach him how to point the answer out. He scored 97th percentile, not enough to be likely to go to our school of choice, but well enough to qualify for the lotteries and maybe even an actual spot. The numbers confused me, though.

The above score breaks down as follows:

The verbal test counts for 1/3 and the non-verbal for 2/3 roughly. So how can 98% and 99% average to 97%?

It’s all because the very few kids who did better in either verbal or non-verbal did much better in those categories. This is an effect of the normalization - they may not have actually done much better in terms of items answered correctly.

Let’s ask a resident mathematician for an example.

Assume you had 100 kids taking the test. In verbal, kid A and kid B were the only ones who did better, but they both got 150 (much better) and 135 non-verbal (only a little worse). Their combined scores are 150 + 2135 = 420. Kid C was the only one who did better in non-verbal, they got a 160 and a 131 verbal (only a little worse) for a combined score of 160 + 2131 = 422. Your combined score was 132 + 2*136 = 404, which was beaten by three kids (A, B, and C), which puts him in in the 97% total.

So, if you combine then normalize the scores the combined score makes more sense. When you normalize then add, it’s much easier for kids to jump ahead by doing well on one test. This is OK with me, because a child that does exceptionally well in one area is more likely to be talented somewhere, as long as they do above average everywhere else.

Now, I don’t believe these tests represent anything other than the ability for a kid to focus on a problem, be well rested, live in a good environment that doesn’t stress them out, and have a level of intelligence appropriate for their age. I don’t think that scoring 99th percentile makes you a genius or says that you were prepped or your parents were rich. These tests are just one way to select children from a very large pool of candidates for a very limited amount of spaces available. I don’t believe that the other choice when you don’t qualify for G&T should be a crappy school. Given our location in NYC we feel lucky not to be in that situation.

Thanks @aaw for a clear explanation of the numbers, quoted “as is” above.