by Teodorescu, Sandra
and Vernic, Raluca
Published in Romanian Journal of Economic Forecasting, 2009, volume 12 issue 4, 82-100
| Requires a PDF viewer such as Xpdf
Prediction is a very important and not so easy task for an actuary. An
insurance company needs predictions of the future claims in order to evaluate premiums, to
assess its financial situation, probabilities of ruin, etc. Therefore, modeling the claimsdistribution is of great importance, but
since this distribution is usually different from the classical ones (e.g. skewed and heavy
tailed), researchers are trying to find new models that can fit better to insurance data.
Such a composite model unifying a Lognormal and a Pareto distribution was introduced by Cooray and Ananda  and generalized by Scollnik . In this paper we go even further and study a composite model obtained from two arbitrary distributions, then exemplify it with the Exponential and Pareto distributions. Some properties and statistical inference are also presented.
composite models, mixture models, Exponential and Pareto
distributions, composite Exponential-Pareto models, parameter estimation