发帖
雷达卡
While this finding can seem intriguing, it is not so surprising when looking at statistics of extreme values. If you take a random sample from a normal distribution, the expected value of the maximum will grow with the sample size. Once your sample size is large enough, however, the expected value of the maximum won’t grow significantly with larger samples. If you consider a standardized normal distribution (which has a mean of 0 and a standard deviation equal to 1), and you take a sample of size 1, by definition the expected value will be 0. With increased sample sizes, one can calculate the expected value of the maximum of the sample. If we were to take a very large sample, the maximum would quite likely be a number of 2 or slightly higher (a number of 2 by definition would be 2 standard deviations above the mean, which in a normal distribution would happen with low probability).
Figure 6.3 shows the expected value of the maximum of such distribution according to the sample size. As you can see from this figure, a sample of size 10 would already produce an expected value close to 1.6, not that far away from a practical maximum of 2. This gives you a hint about why “trying a dozen” can work even when you are sampling candidates from a very large population.
京公网安备 11010802022788号
