The generalized additive model is extended to handle negative binomial resp
onses. The extension is complicated by the fact that the negative binomial
distribution has two parameters and is not in the exponential family. The m
ethodology is applied to data involving DNA adduct counts and smoking varia
bles among ex-smokers with lung cancer. A more detailed investigation is ma
de of the parametric relationship between the number of adducts and years s
ince quitting while retaining a smooth relationship between adducts and the
other covariates.