廣義階層線性模式在強制汽車責任保險理賠資料之應用

Applications of Hierarchical Generalized Linear Models on Compulsory Automobile Liability Insurance Claims Data

楊志強(Chih-Chiang Yang)

《測驗統計年刊》 10期 (2002/12) Pp. 173-195

https://doi.org/10.6773/JRMS.200212.0173

摘要

本研究以廣義階層線性模式分析汽車保險汽車駕駛人於長期縱貫研究下的汽車理賠情形，除了在汽車保險汽車駕駛人出險情形的實質發現之外，主要在於廣義線性階層模式在研究方法學的意義及使用方法，並進一步比較幾個估算方法之優劣及異同。
本研究以民國八十六至八十九年間中華民國強制汽車責任保險汽車駕駛人出險資料為例，這筆資料為特殊變異波以松分配、異質性與同一樣本相依性的特質，並不適合以一般線性迴歸模式分析。因此本研究建議使用以率近似函式及吉氏取樣來估算的廣義階層線性模式，這些方法具有較彈性的模式假設，適合汽車保險汽車駕駛人出險次數的特性。
除了方法學上的研究示範外，本研究並有不同模式間的比較，而且討論了汽車保險理賠與從人因素的關係。

關鍵字

廣義線性階層模式 ； 準近似函式 ； 吉氏取樣 ； 縱貫研究 ； 汽車保險理賠 ； 從人因素

並列摘要

This paper shows both substantive values of applying hierarchical generalized linear model (HGLM) to analyze automobile insurance claims data of a longitudinal study and methodological comparisons of different estimation methods for estimating HGLM. The Taiwan automobile insurance claims dataset, 1997-2000 has the specific extra-Poisson variation, heteroscedasticity, and within-subject dependence characteristics that make it difficult to be analyzed by regular statistical procedures. To investigate appropriately the relationship between the background variables of automobile drivers and the frequency of automobile insurance claims, the researcher used HGLM models by employing quasi-likelihood and Bayesian inference by Gibbs sampling estimation approaches. Both estimation methods that are capable with various model assumptions provide important alternatives to ordinary least square regression.
Results of these analyses verify that the automobile driver's sex, age and marital status play significant roles in the frequency of automobile car insurance claims. Methodological differences between the estimation methods are also outlined.

並列關鍵字

Hierarchical Generalized Linear Model ； Quasi-likelihood ； Gibbs Sampling ； Longitudinal Study ； Automobile Insurance Claims ； Human Factor

參考文獻

Breslow, N.E.,Clayton, D.G.(1993).Approximate inference in generalized linear mixed models.Journal of the American Statistical Association.88,9-25.

Bryk, A.S.,Raudenbush, S.W.(1992).Hierarchical Linear Models: Applications and Data Analysis Methods.CA, Newbury Park:SAGE publications, Inc..

Bryk, A.S.,Raudenbush, S.W.,Congdon, R.T.(1996).Hierarchical Linear Model.Chicago:Scientific Software International, Inc..

Cripe, F.F.,Hunter, R.(1992).Automobile classification systems: age, sex, marital status, and territory are they unfairly discriminatory?.The Actuarial Review.19(3)

Gelman, A.,Rubin, D.B.(1992).Inference from iterative simulation using multiple sequences.Statistical Science.7,457-472.