Premium
Univariate and multivariate mixtures of exponential distributions, with applications in risk modeling
Author(s) -
Cossette Hélène,
Marceau Etienne,
Mtalai Itre,
Veilleux Déry
Publication year - 2021
Publication title -
applied stochastic models in business and industry
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.413
H-Index - 40
eISSN - 1526-4025
pISSN - 1524-1904
DOI - 10.1002/asmb.2606
Subject(s) - univariate , multivariate statistics , mixing (physics) , exponential function , mathematics , statistics , multivariate analysis , econometrics , natural exponential family , statistical physics , exponential distribution , physics , mathematical analysis , quantum mechanics
Mixed exponential distributions are frequently used in actuarial risk modeling. Distributions obtained through mixtures allow greater flexibility in the modeling of nonlife insurance loss amounts. Several research works have studied mixed exponential distributions in univariate and multivariate settings. The present article highlights the usefulness of such distributions and lays the story of the mixing technique behind them. It also explains the underlying link between all these works. We study in detail three univariate and multivariate mixed exponential distributions defined with a discrete mixing random variable (rv). For each of these three univariate (multivariate) mixed exponential distributions and using an appropriate scaling, we identify the continuous mixing rv to which converges in distribution the discrete mixing rv and the corresponding univariate (multivariate) mixed exponential distribution. In a multivariate setting, we show that these three choices of discrete mixing distributions lead us to known Archimedean copulas constructed with continuous mixing rvs. Applications in actuarial science of these distributions are presented throughout the article highlighting their many uses and useful properties.