Happy Learning Hours | Modélisation actuarielle des assurances collectives et participatives
10/05/2022
FAQctuary 2022-2 | Does autocalibration improve goodness of lift?
28/07/2022
Statistical modelling of mortality tables with R
DATE : October 11, 18 & 25, 2022
LENGHT : 3 x 3h30 (from 15:00 to 18h30)
TYPE : Online training
INDUSTRY : Insurance
EXPERTISE : Actuarial
LEVEL : All level
REQUIREMENTS : A computer with a recent version of R studio
LANGUAGE : Speaker: French | Material: English
ACCREDITATION : 9 CPD points | 54 points PPC
PRICE : 525€ excl. VAT
STUDENT PRICE : 90€ excl. VAT
Description
This course focuses on the statistical estimation of static and dynamic mortality models with R. After a general introduction, we review non-parametric (smoothing) approaches and parametric Poisson, Binomial and Gaussian static models. To illustrate this topic, we introduce the R packages MortalityLaws and Demography. In the second Section, we learn how to build mortality experience tables in presence of censorship. These tables make possible the comparison of the mortality experienced by the insurer with the general mortality. The last part of the course is dedicated to dynamic mortality models. We start with the Lee-Carter model and compare the SVD approach to the Poisson model. We present next the CBD and Age-Period-Cohort models and their implementation with the R package STMoMo. The course is concluded with an introduction to multi-population dynamic mortality models. We first focus on the stratified Lee-Carter model and implement it in R. We next present the Li & Lee model and the R package MortalityForecast.
The R code of all illustrations will be distributed to participants and their content will be presented during the lecture.
Target
This module targets an audience of actuaries interested by the practical implementation and a rigorous statistical analysis of mortality tables. The course being self-contained, it is accessible to participants with a mathematical and / or statistical background.
Program
- Introduction
- Survival & death probabilities
- Reading datasets from HMD to R
- Static mortality probabilities
- Non-parametric static smoothing
- Parametric static smoothing
- Experience mortality tables with censorship
- Prospective mortality tables
- Multi-population prospective mortality tables
Acquired skills
At the end of this training, the participants will be able
- To have a global viewpoint of features of mortality data available on the web;
- To estimate in R any static parametric models with a rigorous statistical method and to avoid common pitfalls;
- To construct an experience mortality curve and to manage censorship;
- To compute prospective mortality tables with age, period and cohort models for a single population;
- To use R packages, MortalityLaws, Demography, StMoMo, MortalityForecast;
- To estimate multi-populations models;
- To write her/his own code for implementing non-standard mortality models.
This training is given by
Donatien Hainaut
Scientific Director
About the trainer
Donatien is a Qualified Actuary and holds a PhD in the area of Assets and Liability Management. He is Scientific Director at Detralytics and professor at UCLouvain where he is Director of the new Master program in Data Science, statistical orientation.
Prior to this he held several positions as associate professor at Rennes School of Business and the ENSAE in Paris. He also has several field experiences having worked as Risk Officer, Quantitative Analyst and ALM Officer.
