The GAM (Hastie and Tibshirani 1990) postulates that the response is additively related to the independent variables via the equation
Because of the flexibility provided by allowing the [f.sub.i] to be nonparametric functions, as well as the fact that it is easy to fit these models using standard statistical software such as S-Plus, the GAM has gained favor as a powerful analytical tool.
The term "concurvity" refers to the presence of a specific type of (approximate) additive dependence among the independent variables in a GAM. For the general GAM in Equation 1, an additive dependence of the form
This implies that concurvity is highly dependent on the particular GAM being considered.
An exact concurvity relation causes the GAM to be unidentifiable (Ramsay et al.