Graphical Diagnostic for Mortality Data Modeling
Preprint
- 18 July 2016
- preprint
- Published by MDPI AG in Preprints.org
Abstract
The main contribution of this paper is to develop a graphical tool to evaluate the suitability of a candidate parametric model to fit a data set. The practical motivation is to check the adequacy of the so called SAINT model proposed in Jarner and Kryger (2011). This model has been implemented in practice by an important European pension fund concerned with setting annuity reserves for all current or former employees of Denmark. So, one particular relevant question is whether this mortality model is actually fitting old-age. Our graphical test can be described as follows. First we transform the data by means of the parametric model which is being evaluated. Should the model be correct, the transformed data will be in accordance with an Exponential distribution with rate 1. Then we construct a family of local linear hazard estimates based on the data on the transformed scale. Finally we use the statistical tool SiZer to assess the goodness-of-fit of the Exponential distribution to the data on the transformed scale. If the model is correct the SiZer map should not reveal any significant feature in the family of kernel estimates. Our method allow us to establish a diagnostic regarding the validity of the SAINT model when describing mortality patterns in four different countries.Keywords
All Related Versions
This publication has 16 references indexed in Scilit:
- Long-range dependence analysis of Internet trafficJournal of Applied Statistics, 2010
- Local linear hazard rate estimation and bandwidth selectionAnnals of the Institute of Statistical Mathematics, 2010
- Local linear density estimation for filtered survival data, with bias correctionStatistics, 2009
- SiZer analysis for the comparison of regression curvesComputational Statistics & Data Analysis, 2008
- SiZer for time series: A new approach to the analysis of trendsElectronic Journal of Statistics, 2007
- Dependent SiZer: Goodness-of-Fit Tests for Time Series ModelsJournal of Applied Statistics, 2004
- SiZer for Exploration of Structures in CurvesJournal of the American Statistical Association, 1999
- Analysis of oldest-old mortality: lifetables revisitedThe Annals of Statistics, 1998
- Equivalence of Smoothing Parameter Selectors in Density and Intensity EstimationJournal of the American Statistical Association, 1988
- A Kernel Method for Smoothing Point Process DataJournal of the Royal Statistical Society Series C: Applied Statistics, 1985