Loss modeling using Burr mixtures
The first-ever real data application of a two-component Burr mixture distribution is provided. It is fitted to three loss data sets: fire loss claims in Denmark, fire loss claims for three building categories in Belgium and fire loss data in Norway. Each of these data sets exhibits significant bimod...
Saved in:
Main Authors: | , , |
---|---|
Format: | Article |
Published: |
Springer
2018
|
Subjects: | |
Online Access: | http://eprints.um.edu.my/22714/ https://doi.org/10.1007/s00181-017-1269-7 |
Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
Summary: | The first-ever real data application of a two-component Burr mixture distribution is provided. It is fitted to three loss data sets: fire loss claims in Denmark, fire loss claims for three building categories in Belgium and fire loss data in Norway. Each of these data sets exhibits significant bimodality. The fits of the two-component Burr mixture distribution are compared to those of five other two-component mixture distributions: the two-component Weibull mixture, two-component gamma mixture, two-component Pareto mixture, two-component lognormal mixture and the two-component exponential mixture distributions. The Burr mixture distribution is shown to give the best fit for each data set. The relative performances of the fitted distributions were assessed in terms of Akaike information criterion values, Bayesian information criterion values, consistent Akaike information criterion values, corrected Akaike information criterion values, Hannan–Quinn criterion values, density plots and probability–probability plots. |
---|