On the statistical differences between binary forecasts and real-world payoffs

We map the difference between (univariate) binary predictions, bets and “beliefs” (expressed as a specific “event” will happen/will not happen) and real-world continuous payoffs (numerical benefits/harm from an event) and show the effect of their conflation and mischaracterization in the decision-science literature. We also examine the differences under thin and fat tails. The effects: [A] Spuriousness of many psychological results, particularly those documenting that humans overestimate tail probabilities. We quantify such conflations. [B] Being a “good forecaster” in binary space doesn’t lead to having a good actual performance, and vice versa, especially under nonlinearities. A binary forecasting record is likely to be a reverse indicator under some classes of distributions or deeper uncertainty. [C] Machine Learning: Some nonlinear payoff functions, while not lending themselves to verbalistic expressions, are well captured by ML or expressed in option contracts. Fattailedness: The difference is exacerbated in the power law classes of probability distributions.     

Model risk in credit risk

We provide sharp analytical upper and lower bounds for value‐at‐risk (VaR) and sharp bounds for expected shortfall (ES) of portfolios of any dimension subject to default risk. To do so, the main methodological contribution of the paper consists in analytically finding the convex hull generators for the class of exchangeable Bernoulli variables with given mean and for the class of exchangeable Bernoulli variables with given mean and correlation in any dimension. Using these analytical results, we first describe all possible dependence structures for default, in the class of finite sequences of exchangeable Bernoulli random variables. We then measure how model risk affects VaR and ES.     

Compound trend renewal process with discounted claims: a unified approach

We derive recursive formulas for the moments of compound trend renewal sums with discounted claims. An integral expression for the moment generating function of this risk process is then obtained, from which particular distribution functions are found. We extend the compound (deterministic) trend renewal process by assuming a stochastic trend, a stochastic force of net interest and a stochastic dependence between the inter-occurrence times and the severities of the claims. Finally, stochastic dominance ordering is also observed between the compound trend renewal process and an associated non-homogeneous Poisson process.     

Estimation of P[Y < X] for generalized exponential distribution

This paper deals with the estimation of P[Y < X] when X and Y are two independent generalized exponential distributions with different shape parameters but having the same scale parameters. The maximum likelihood estimator and its asymptotic distribution is obtained. The asymptotic distribution is used to construct an asymptotic confidence interval of P[Y < X]. Assuming that the common scale parameter is known, the maximum likelihood estimator, uniformly minimum variance unbiased estimator and Bayes estimator of P[Y < X] are obtained. Different confidence intervals are proposed. Monte Carlo simulations are performed to compare the different proposed methods. Analysis of a simulated data set has also been presented for illustrative purposes.Part of the work was supported by a grant from the Natural Sciences and Engineering Research Council     

Minimax estimation under convex loss when the parameter interval is bounded

In the present paper families of truncated distributions with a Lebesgue density forx=(x ₁,...,x _n ) ε ℝ ⁿ are considered, wheref ₀:ℝ → (0, ∞) is a known continuous function andC _n (ϑ) denotes a normalization constant. The unknown truncation parameterϑ which is assumed to belong to a bounded parameter intervalΘ=[0,d] is to be estimated under a convex loss function. It is studied whether a two point prior and a corresponding Bayes estimator form a saddle point when the parameter interval is sufficiently small.     

Portfolio optimization under a generalized hyperbolic skewed t distribution and exponential utility

In this paper, we show that if asset returns follow a generalized hyperbolic skewed t distribution, the investor has an exponential utility function and a riskless asset is available, the optimal portfolio weights can be found either in closed form or using a successive approximation scheme. We also derive lower bounds for the certainty equivalent return generated by the optimal portfolios. Finally, we present a study of the performance of mean–variance analysis and Taylor’s series expected utility expansion (up to the fourth moment) to compute optimal portfolios in this framework.     

Parisian types of ruin probabilities for a class of dependent risk-reserve processes

For a rather general class of risk-reserve processes, we provide an exact method for calculating different kinds of ruin probabilities, with particular emphasis on variations over Parisian type of ruin. The risk-reserve processes under consideration have, in general, dependent phase-type distributed claim sizes and inter-arrivals times, whereas the movement between claims can either be linear or follow a Brownian motion with linear drift. For such processes, we provide explicit formulae for classical, Parisian and cumulative Parisian types of ruin (for both finite and infinite time horizons) when the clocks are phase-type distributed. An erlangization scheme provides an efficient algorithmic methods for calculating the aforementioned ruin probabilities with deterministic clocks. Special attention is drawn to the construction of specific dependency structures, and we provide a number of numerical examples to study its effect on probabilities.     

On a property of the sequential t-test

Abstract

1. Introduction. In one of his papers [1], and later in his book on sequential analysis [2], Wald introduced a general method for constructing sequential tests of composite hypotheses and applied the method to construct a sequential t-test. Since Wald devoted a considerable amount of space and mathematics on his t-test, it has been taken for granted that he proved certain optimum properties of the test. It is the purpose of this note to show that the test cannot possess one of the properties thought to hold for it.     

BAYESIAN ANALYSIS OF AGGREGATE LOSS MODELS

This paper describes a Bayesian approach to make inference for aggregate loss models in the insurance framework. A semiparametric model based on Coxian distributions is proposed for the approximation of both the interarrival time between claims and the claim size distributions. A Bayesian density estimation approach for the Coxian distribution is implemented using reversible jump Markov Chain Monte Carlo (MCMC) methods. The family of Coxian distributions is a very flexible mixture model that can capture the special features frequently observed in insurance claims. Furthermore, given the proposed Coxian approximation, it is possible to obtain closed expressions of the Laplace transforms of the total claim count and the total claim amount random variables. These properties allow us to obtain Bayesian estimations of the distributions of the number of claims and the total claim amount in a future time period, their main characteristics and credible intervals. The possibility of applying deductibles and maximum limits is also analyzed. The methodology is illustrated with a real data set provided by the insurance department of an international commercial company.