Example 12: Mixtures and Assessing Models vs. Large
Loss Experience
Large
Loss Listing
Exposure Distribution
Curve
Percent Total
Total
1.0
Losses XS
Policy Limit
NumL2
Click compute to see results...
Notes
This example is to help you assess whether large loss
experience comes from the severity curves you would expect, given its class.
It is intended as a "quick and dirty" review, not as a substitute for more
rigorous statistical analysis.
Input a list of large losses excess some censoring
point (the "Losses XS" value).
Input policy limit, or truncation point. If left
blank, the largest loss is used.
Input the length of the numerical approximation to
use (actual length is 2^NumL2, so 7 correspods to 128 buckets).
Select the mixture weights from the five drop downs,
showing the available distributions. Weights should sum to one.
Model shows a comparison of the density (f),
distribution (F) and survival functions; the expected excess value and a QQ
plot. The excess values are computed numerically from a sample, so they tend
to zero.
All curves in the public version of this model are identical. See the Appendix
for instruction on how to create your own parameter file, without which this
example is practically useless. The dummy curve is too thick tailed to be
representative of the data shown.
Examples include losses similar to those from a large fleet of heavy trucks and
losses similar to those from a large fleet of private passenger and light
vehicles.
References
Klugman, S., H. Panjer and G. Willmot, "Loss Models
From Data to Decisions" Wiley 1998
Hogg R. and S. Klugman, "Loss Distributions" Wiley 1984