Please read this excellent commentary by Nate Silver, and let me follow up on his comment about the 2016 election. In statistical analysis it's is extremely important to understand the distinction between probabilities and odds. Because probabilities are bounded by 0 in 1, increases in the probability a candidate will win are non-linear, and consequently the meaning of these increases is not intuitive. For example on the night of the 2016 election silver predicted Clinton had a 70% chance of winning. When this means the odds of a Clinton win or 2.33. In other words if you were picking from 3 straws a little more than two of them would be for Clinton and one would be for Trump. Those odds should have terrified us, but we were fixated on the 70% and not doing the math (odds = p/1-p).
No comments:
Post a Comment