GAM is designed to maximize the prediction of a dependent variable, y, from various distributions by estimating the nonparametric functions of the predictors that link to the dependent variable through a link function. The notion of GAM is, where an exponential family, E, is specified for y, along with the g link function; f denotes the link function of predictors.

The gam function is contained in the mgcv package, so, install this package first and load it into an R session. Next, load the Boston dataset (Housing Values in the Suburbs of Boston) from the MASS package. From the dataset, we use dis (the weighted mean of the distance to five Boston employment centers) as the dependent variable, and nox (nitrogen oxide concentration) as the independent variable, and then input them into the gam function to generate a fitted model.

Similar to glm, gam allows users to summarize the gam fit. From the summary, one can find the parametric parameter, significance of smoothed terms, and other useful information.