Hacker News Reader

for example, if you wanted to do research on whether or not smarter CEOs increase shareholder value, how would you go about it ? Intelligence is hard to measure, but we know things like test scores correlate with it to some degree. So, maybe, you would see how well companies do against CEO SAT scores or undergraduate institution they attended or their high school gpa

View on HN

Then you have to admit that measuring correlation between "smarter CEOs" and "shareholder value" is impossible and measure correlation between "CEO SAT scores" and "shareholder value" instead. Yeah, not that click-baity, but that would be a real study, unlike this one.

if you have a correlation between A and B, and a correlation between B and C, then A and C are correlated.

This is not true in general.

Consider two independent fair die rolls, with A and C being the outcomes. Let B be the sum of the outcomes. The higher A is, the higher B is likely to be, so A and B are positively correlated. So are B and C. However there is no correlation between A and C as they are independent.

Oh. True, thanks for the explanation :)

Reply on news.ycombinator.com