Bayesian Inference: Updating Beliefs with New Evidence
Summary
This beginner's guide explains Bayesian statistics as a framework for updating uncertainty when new evidence arrives. It contrasts Bayesian probability, interpreted as confidence in possible outcomes, with the frequentist view of probability as long-run frequency across repeated trials. Using coin flips and an election as illustrations, it shows how people can begin with different prior beliefs and revise them as evidence accumulates.
The article derives Bayes' rule from conditional probability and applies the idea to estimating the probability that a coin lands heads. A Bernoulli model describes each flip, while a Beta distribution represents uncertainty about the coin's head probability; as observations accumulate, the example's posterior becomes more concentrated. This is an educational illustration, not a market forecasting study. The supplied excerpt omits part of the worked calculation, and the guide defers a fuller treatment of conjugate priors. It also does not compare Bayesian estimates against frequentist methods on trading data.
Key ideas
- Bayesian inference updates prior beliefs using observed evidence to form posterior beliefs.
- Frequentist probability describes long-run event frequencies, while Bayesian probability represents uncertainty about outcomes.
- Bayes' rule follows from the definition of conditional probability.
- A Bernoulli model can represent coin flips with a fixed probability of heads.
- In the coin example, a Beta posterior concentrates as more flip data is observed.
Tags
This summary was written by Stratmill's research agent from the original; it is not a copy of the source.