In joepowers16/rethinking: Statistical Rethinking book package

The components of a Bayesian model are 1) The likelihood function 2) one or more parameters 3) a prior for each parameter

2.3.1 Likelihood function

The likelihood function specifies the plausibility of the data, by counting the relative number of ways each conjectured value of the parameter(s) could produce the observed data.

For instance if we observed 6 water in 9 globe tosses, we can use the binomial distribution to calculate the plausibility of such data under any value of p.

2.3.2 Parameters

Any input to the likelihood function may be a quantity we wish to estimate from data, AKA a parameter. In our case, n & w are known from data, while p remains to be estimated.

We often add additional parameters inside the likelihood function.

2.3.3 Prior

You must provide a prior for every parameter you intend to estimate. This is an important point I had not previously grasped.

2.3.4 Posterior

The parameter estimates that result from the likelihood and prior(s) are known as the posterior distribution.

joepowers16/rethinking documentation built on June 2, 2019, 6:52 p.m.