R in a Nutshell

Chapter 20. Regression Models

A regression model shows how a continuous value (called the response variable, or the dependent variable) is related to a set of other values (called the predictors, stimulus variables, or independent variables). Often, a regression model is used to predict values where they are unknown. For example, warfarin is a drug commonly used as a blood thinner or anticoagulant. A doctor might use a regression model to predict the correct dose of warfarin to give a patient based on several known variables about the patient (such as the patient’s weight). Another example of a regression model might be for marketing financial products. An analyst might estimate the average balance of a credit card customer (which, in turn, affects the expected revenue from that customer).

Sometimes, a regression model is simply used to explain a phenomenon, but not to actually predict values. For example, a scientist might suspect that weight is correlated to consumption of certain types of foods, but wants to adjust for a variety of factors, including age, exercise, genetics (and, hopefully, other factors). The scientist could use a regression model to help show the relationship between weight and food consumed by including other variables in the regression. Models can be used for many other purposes, including visualizing trends, analysis of variance tests, and testing variable significance.

This chapter looks at regression models in R; classification models are covered in Chapter 21 ...