@statistical-rethinking @book.

09/02/23 19:03:23

Chapter 3

The models work is done after calculating the posterior distribution, but no, we’ve to interpret it.

3.2

Sampling can be done to summarise the posterior. Review notion of a loss function

3.3

  • A likelihood function works both ways. Given a realized observation, the likelihood function tells us how plausible that observation is. And given only parameters the likelihood function gives us a distribution of possible observations that we can sample from.(?)
  • Bayesian models are always generative.

It gets a bit nested when it comes to generating distributions for distributions.