So far we have been fitting lines to points.

- We haven't really been doing statistics because we haven't really introduced any probabilities.
- We don't really have a
**model**. - And we'll need a model if we are going to perform inference:
- to get a sense of estimation uncertainty,
- to test for the usefulness of explanatory variables (\(x\)),
- and to put "error bars" on our predictions from the line(s).