Kim, Minkun, Bezbradica, Marija ORCID: 0000-0001-9366-5113 and Crane, Martin ORCID: 0000-0001-7598-3126 (2024) Bayesian Hierarchical Risk Premium Modeling with Model Risk: Addressing Non-Differential Berkson Error. Applied Sciences, 15 (210). pp. 1-38. ISSN 2076-3417
For general insurance pricing, aligning losses with accurate premiums is crucial for insurance companies’ competitiveness. Traditional actuarial models often face challenges like data heterogeneity and mismeasured covariates, leading to misspecification bias. This paper addresses these issues from a Bayesian perspective, exploring connections between Bayesian hierarchical modeling, partial pooling techniques, and the Gustafson correction method for mismeasured covariates. We focus on Non-Differential Berkson (NDB) mismeasurement and propose an approach that corrects such errors without relying on gold standard data. We discover the unique prior knowledge regarding the variance
of the NDB errors, and utilize it to adjust the biased parameter estimates built upon the NDB covariate. Using simulated datasets developed with varying error rate scenarios, we demonstrate the superiority of Bayesian methods in correcting parameter estimates. However, our modeling process highlights the challenge in accurately identifying the
variance of NDB errors. This emphasizes the need for a thorough sensitivity analysis of the relationship between our prior knowledge of NDB error variance and varying error rate scenarios.
Item Type: | Article (Published) |
---|---|
Refereed: | Yes |
Uncontrolled Keywords: | Bayesian hierarchical model; heterogeneity; non-differential Berkson measurement error; aggregate insurance claim; risk premium; partial pooling; Gustafson correction |
Subjects: | Computer Science > Computer engineering Computer Science > Computer security Computer Science > Software engineering |
DCU Faculties and Centres: | DCU Faculties and Schools > Faculty of Engineering and Computing DCU Faculties and Schools > Faculty of Engineering and Computing > School of Computing Research Institutes and Centres > ADAPT |
Publisher: | MDPI AG |
Official URL: | https://www.mdpi.com/2076-3417/15/1/210 |
Copyright Information: | Authors |
ID Code: | 30668 |
Deposited On: | 17 Jan 2025 14:40 by Gordon Kennedy . Last Modified 17 Jan 2025 14:40 |
Full text available as:
Preview |
PDF
- Requires a PDF viewer such as GSview, Xpdf or Adobe Acrobat Reader
Creative Commons: Attribution 4.0 9MB |
Dimensions Badge
Downloads
Downloads per month over past year
Archive Staff Only: edit this record