Workshop on Generalized Linear Model and Its Application in General Insurance
Brief Information
General Description
Generalized linear models have been in use for many years by actuaries, and there is no shortage of textbooks and scholarly articles on their underlying theory and application in solving any number of useful problems. Actuaries have for many years used GLMs to classify risks, but it is only relatively recently that levels of interest and rates of adoption have increased to the point where it now seems as though they are near-ubiquitous. Our ultimate goal is to give the participants all of the additional tools he or she needs to build a market-ready classification plan from raw premium and loss data.
Instructor
Prof. David Landriault, FSA, FCIA, Ph.D. (University of Waterloo, Canada)
Aim
To introduce the GLM and their applications in Non-Life Insurance context for both of practitioners and Actuarial Science lecturer in Indonesia.
Level of audience
Beginner – Novice
Duration
2 days with coffee breaks and lunches, approximately 18 hours
Method of delivery
Presentation (75%) and Computer programming session (25%) (please describe your favorite Stat. Software but preferably R)
Topics
(included but not limited to):
- Insurance data (Data exploration, type of variables, assessing distributions)
- Introduction to GLM
- Exponential family distribution with few examples
- (Log) likelihood, Score function/vector, Information function/Matrix
- Short explanation about Root finder algorithm, Newton-Raphson, Fisher Scoring, IRWLS.
- Model for count data (Poisson regression, Poisson overdispersion, Quasi-likelihood)
- Categorical responses (Binary responses, Logistic regression)
- Continuous responses (Gamma, Inverse Gaussian, Tweedie)
Participants
Some of the expected participants are not only from academics but also from life and general insurance industry practitioners. There are only 20 seats available.
Schedule
Registration
To participate the workshop, please register through The SEAMS-UGM 2019 Conference’s website, or just click the following button.