The paper's conclusion that a can of coke per day is associated with a 'relative risk' of 1.18. That means X% -> x*(1.18)%, where X is whatever the baseline ris

The paper's conclusion that a can of coke per day is associated with a 'relative risk' of 1.18. That means X% -> x*(1.18)%, where X is whatever the baseline risk is. The blog introduces the result as increasing 'the likelihood of developing diabetes by 18%', which means the same thing. It then incorrectly translates this to a 'marginal chance of 18%' -- which means X% -> X+18%. And the math uses that. X%->X+18% is 'a marginal risk of 18%', or 'an increase of 18 percentage points' or 'a 0.18 increase in absolute risk'.