Scott Alexander explores how summary statistics can be misleading when describing relationships between variables, using examples of IQ's correlation with crime and income.
Longer summary
Scott Alexander discusses two examples where summary statistics can be misleading. The first example involves the relationship between IQ and crime, where a large difference in average IQ between offenders and non-offenders coexists with a low correlation coefficient. The second example concerns the relationship between IQ and income, where substantial differences in average income across IQ deciles coexist with a relatively low correlation coefficient. In both cases, Scott emphasizes the importance of looking beyond summary statistics and considering the full distribution of data, potentially by examining scatter plots.
Shorter summary