![[statistics#goals of statistics]]
The task of **statistical inference** is to infer properties of the data distribution ("data generating process").
Mostly via [[point estimator]] / [[interval estimator]] of the [[parameter]] and via [[hypothesis test]].
There are both [[Bayes rule|Bayesian]] and [[frequentist]] ways of doing this.
See [[statistical decision theory]] for how to compare different methods at doing different things.
The d.g.p. can often be described by a [[Bayesian network]].
This is typically [[unsupervised]];
Sometimes the data can be factored into [[covariate]]s $X$ and the [[response]] $Y$ in which case we might want to infer properties of the conditional distribution $Y \mid X$ (eg the [[conditional expectation]]) via [[supervised]] learning. eg [[regression]]
# sources
[[STAT 111]] chap 4.3
[[STAT 111]] 2.3.4