Hacker News new | past | comments | ask | show | jobs | submit login

In contrast with frequentism, in Bayesianism it makes perfect sense to attach probabilities to one-shot events, such as who will win the next election, or even past events that we’re unsure about.

Does frequentism really require actually performing the experiment? Or is imagining doing the experiment good enough? I would say

  »Candidate X will win the next election with a probability of Y percent.«
is just a shorthand for

  »The following sets of states and possible evolutions of those states are
  compatible with my knowledge about the world and in Y percent of the cases
  candidate X wins the next election.«
which seems not to different from a coin flip where the different outcomes are also due to imperfect knowledge of the initial state. The difference is that it is easy to sample the set of initial states for a coin flip by just repeatedly flipping a coin from slightly different initial states due to human imperfections in doing this task. Sampling the initial states of an election in the same way is obviously not possible and I have admittedly no real clue how people arrive at a meaningful number in practice. A similar example seems to be the probability of rain at some place some time into the future in which case it is possible to sample the set of initial states by running a weather model repeatedly.

I interpret it as a reference system. To a frequentist, a probabilistic statement is "this coin flip is 50% heads in reference to this set of coin flips." Your elements of the statement are the event, the probability, and the reference set of events.

Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | Legal | Apply to YC | Contact