Regression is a supervised learning method, which is employed to model and analyze the relationship between a dependent (response) variable and one or more independent (predictor) variables. One can use regression to build a prediction model, which can first be used to find the best fitted model with minimum squared errors of the fitted values. The fitted model can then be further applied to data for continuous value predictions.

There are many types of regression. If there is only one predictor variable, and the relationship between the response variable and independent variable is linear, we can apply a linear model. However, if there is more than one predictor variable, a multiple linear regression method should be used. When the relationship is nonlinear, one can use a nonlinear model to model the relationship between the predictor and response variables.

In this chapter, we will introduce how to fit a linear model into data with the lm function. Next, for distribution in other than the normal Gaussian model (for example, Poisson or Binomial), we use the glm function with an appropriate link function correspondent to the data distribution. Finally, we will cover how to fit a generalized additive model into data using the gam function.