D-Splines: Estimating Rate Schedules Using High-Dimensional Splines With Empirical Demographic Penalties

Carl P. Schmertmann , Florida State University

I develop and test a new class of penalized spline models for estimating demographic rate schedules, with a particular interest in applying the methods to sparse data from small areas or small sub-populations. I propose adaptations of the P-spline approach (called D-splines) that regularize and smooth high-dimensional splines by using specific demographic knowledge rather than generic arithmetical rules. Preliminary tests of four alternative D-spline estimators on simulated small-area mortality data are promising: D-spline estimators appear to have low errors and to produce schedules that reliably reflect known properties of human mortality schedules.

See extended abstract

 Presented in Session 185. Statistical Demography