Bayesian Predictive Modeling for Exponential-Pareto Composite Distribution

By M. S. Aminzade, Min Deng

Download PDF of Full Text

Abstract

Composite distributions have well-known applications in the insurance industry. In this paper, a composite exponential-Pareto distribution is considered, and the Bayes estimator under the squared error loss function is derived for the parameter q, which is the boundary point for the supports of the two distributions. A predictive density is constructed under an inverse gamma prior distribution for the parameter q, and the density is used to estimate the value at risk (VaR). Goodness of fit of the composite model is verified for a generated data set. The accuracy of the Bayes and VaR estimates is assessed via simulation studies. The “best” value for hyperparameters of the inverse gamma prior distribution are found via an upper bound on the variance of the prior distribution. Simulation studies indicate that when the “best” values of hyperparameters are used in the Bayes estimator, the estimator is consistently more accurate than maximum likelihood estimation.

Keywords: Maximum likelihood estimate, inverse gamma, Bayes estimate, exponential-Pareto composite density, predictive density, value at risk

Citation

Aminzade, M. S., and Min Deng, "Bayesian Predictive Modeling for Exponential-Pareto Composite Distribution," Variance 12:1, 2018, pp. 59-68.

Taxonomy Classifications


Subscribe to the RSS Feed

Email List

Sign up today for the Variance e-mail list and receive updates about new issues, articles, and special features.

Mission Statement

Variance (ISSN 1940-6452) is a peer-reviewed journal published by the Casualty Actuarial Society to disseminate work of interest to casualty actuaries worldwide. The focus of Variance is original practical and theoretical research in casualty actuarial science. Significant survey or similar articles are also considered for publication. Membership in the Casualty Actuarial Society is not a prerequisite for submitting papers to the journal and submissions by non-CAS members is encouraged.