|
|
|
|
|
|
by nostrademons
2805 days ago
|
|
|
His rebuttal to the "IQ has fat tails" argument isn't very convincing. He's arguing that IQ should be defined as your percentile rank converted into a numeric score based on a Gaussian of mean 100 and stddev 15, but then goes on to say that yes, the results of most IQ tests will deviate from that Gaussian in the extremes, but that's because the tests are inaccurate, and will have whatever distribution the test-makers give them (which is probably not Gaussian). However, a definition that differs from basically every empirical measurement of that quantity isn't a very good definition! That's like defining pi as 22/7 or the fine structure constant as 1/137 because it makes the math easier. Sure, that's true, and it makes it easier to explain to laypeople...but you're going to have problems when you try to predict or draw conclusions about the results of any experiments. Same with IQ - you can define it as the percentile rank converted to a scale with 100 as the mean and 15 as the stddev, but if that yields the prediction that children with a 175 IQ (5 stddev) are going to be 1 in 2 million while the observed frequency on say a Stanford Binet L-M is 1 in 10,000, you're going to draw drastically erroneous conclusions. (In practice the WISC3 and Stanford-Binet 5th ed aren't very useful for highly-gifted children anyway, because they ceiling out around 140 IQ, and basically all IQs over that are estimated with either the Binet L-M or early administration of the SATs.) |
|
|