Analysis of Mortality Data using Smoothing Spline Poisson Regression

We study a smoothing spline Poisson regression model for the analysis of mortality data. Being a non-parametric approach it is intrinsically robust, that it is a penalized likelihood estimation method makes available an approximate Bayesian confidence interval and importantly the software gss, its implementation on the freely available statistical package R, makes it easily accessible to the user. All of this make it an attractive alternative to (usually computationally intensive) fully Bayesian analysis while avoiding the complexity of high-dimensional prior specification.