6 Background on regression modelling
6.1 Summary
At a pure mathematical level, the methods described in this book have two purposes:
- prediction (regression to predict an outcome variable, or more precisely the distribution of the outcome, given some set of inputs);
- comparison (comparing predictions for different values of the inputs, to make simple comparisons between groups, or to estimate causal effects (see chapters 18-21).
In this chapter:
- the favored technique of fake-data simulation is used to understand a simple regression;
- with using an example (height and earnings) warning agains unwarranted causal interpretations;
- historical origins of regression are discussed ;