Approximate Bayesian Computation推荐.pdfVIP

CHAPTER 10 Approximate Bayesian Computation The Variability Hypothesis I have a soft spot for crank science. Recently I visited Norumbega Tower, which is an enduring monument to the crackpot theories of Eben Norton Horsford, inventor of double-acting baking powder and fake history. But that’s not what this chapter is about. This chapter is about the Variability Hypothesis, which “originated in the early nineteenth century with Johann Meckel, who argued that males have a greater range of ability than females, especially in intelligence. In other words, he believed that most geniuses and most mentally retarded people are men. Because he considered males to be the ’superior animal,’ Meckel concluded that females’ lack of var‐ iation was a sign of inferiority.” From /wiki/Variability_hypothesis. I particularly like that last part, because I suspect that if it turns out that women are actually more variable, Meckel would take that as a sign of inferiority, too. Anyway, you will not be surprised to hear that the evidence for the Variability Hypothesis is weak. Nevertheless, it came up in my class recently when we looked at data from the CDC’s Behavioral Risk Factor Surveillance System (BRFSS), specifically the self-reported heights of adult American men and women. The dataset includes responses from 154407 men and 254722 women. Here’s what we found: • The average height for men is 178 cm; the average height for women is 163 cm. So men are taller, on average. No surprise there. • For men the standard deviation is 7.7 cm; for women it is 7.3 cm. So in absolute terms, men’s heights are more variable. • But to compare variability between groups, it is more meaningful to use the coef‐ ficient of variation (CV), which is the standard deviation divided by the m