Bounds for Ratios of Posterior Expectations: Applications in the Collective Risk Model (Q4780928)
From MaRDI portal
!
This is the item page for this Wikibase entity, intended for internal use and editing purposes. Please use the normal view instead:
scientific article; zbMATH DE number 1833673
| Language | Label | Description | Also known as |
|---|---|---|---|
| default for all languages | No label defined |
||
| English | Bounds for Ratios of Posterior Expectations: Applications in the Collective Risk Model |
scientific article; zbMATH DE number 1833673 |
Statements
Bounds for Ratios of Posterior Expectations: Applications in the Collective Risk Model (English)
0 references
21 November 2002
0 references
collective risk model
0 references
variance premium principles
0 references
Bayesian robustness
0 references
Esscher premium principle
0 references
0 references
0.8811728
0 references
0.8654299
0 references
0.85886484
0 references
0.8580375
0 references
0.8573406
0 references
0 references
In the collective risk model a risk is a sequence \(N, X_1,X_2,\ldots\) of random variables, where \(N\) is the number of claims and \(X_{i}\) is the size of the \(i\)-th claim, \(i=1,2,\ldots\), the random variables \(N\) and \(X_{i}, \forall i\) are independent, and \(X_{i}\) are i.i.d. random variables. The aggregate loss probability density function of the risk is given by \(f(x|\theta)=\sum_{n=0}^{\infty}f^{*n}(x)p(n|\theta)\), where \(f(x)\) is the common probability function of \(X_{i}\); \(p(n|\theta)= \text{Prob}[N=n]\), \(\theta\in\Theta\) is an unknown structural parameter and \(f^{*n}(x)\) is the \(n\)-fold convolution of the corresponding probability function. The authors obtain bounds for the Bayesian premium which is defined as a real number \(P^{*}_{\pi_0}(m)\) minimizing the posterior expected loss \(E_{\pi_0(\theta|m)}[L(P(\theta),P^{*}_{\pi_0}(m)]\), where \(\pi_0(\theta)\) denotes the prior density function of \(\theta\); \(m\) is the observed realization of a claim experience (in a period of time \(t\) \(m\) claims are observed); \(\pi_0(\theta|m)\) denotes the posterior density function of \(\theta\); \(L(x,P)=x(x-P)^2\) is used for the variance premium principle and \(L(x,P)=e^{\alpha x}(x-P)^2\) is used for the Esscher premium principle. For modeling the prior uncertainties the authors use the \(\varepsilon\)-contaminated class of distributions \(\Gamma_{\varepsilon}= \{\pi(\theta)=(1-\varepsilon)\pi_0(\theta)+\varepsilon q(\theta)|q\in {\mathbf P}\}\), where \({\mathbf P}\) is a set of probability distributions reflecting the type of deviations from \(\pi_0\).
0 references