Danish Fire Insurance Claims

Source Notebook

Fire insurance claims in Denmark from Thursday 3rd January 1980 until Monday 31st December 1990

Details

These data are used in the book Modelling Extremal Events by Embrechts, Kluppelberg and Mikosch to illustrate modelling extremal events with the Pareto distribution.

Examples

Basic Examples (1) 

In[1]:=
ts = ResourceData["Danish Fire Insurance Claims"]
Out[1]=

Plot the data:

In[2]:=
dataDanish = ts["Values"];
ListPlot[dataDanish, Sequence[
 Filling -> Axis, PlotRange -> All, Axes -> False, Frame -> True, FrameLabel -> {"Time Index", "Claim Size"}, ImageSize -> 500, BaseStyle -> {FontSize -> 14}]]
Out[3]=

Visualizations (1) 

Construct a histogram of the log claim size:

In[4]:=
ts = ResourceData["Danish Fire Insurance Claims"];
dataDanish = ts["Values"];
Histogram[
 Log[dataDanish], Sequence[{0, 4, 0.05}, PlotRange -> All, Axes -> False, Frame -> True, FrameLabel -> {"\!\(\*SubscriptBox[\(log\), \(e\)]\)(Claim Size)", "Count"}, ImageSize -> 500, BaseStyle -> {FontSize -> 14}]]
Out[6]=

Analysis (2) 

Threshold the data and fit to a Pareto distribution:

In[7]:=
u = 10.0;
ts = ResourceData["Danish Fire Insurance Claims"];
dataDanish = ts["Values"];
data = Sort@Select[dataDanish, # > u &];
Nd = Length[data];
Clear[k, \[Alpha], \[Mu]];
d = EstimatedDistribution[data, ParetoDistribution[k, \[Alpha], \[Mu]]]
Out[13]=

Compare empirical and theoretical cumulative probability distributions:

In[14]:=
modelCDF = Table[{Log[x], CDF[d, x]}, {x, Min[data], Max[data], 0.1}];
empiricalCDF = Transpose[{Log[data], N@Table[n/Nd, {n, 1, Nd}]}];
ListPlot[empiricalCDF, Epilog -> {Line[modelCDF]}, Sequence[
 Axes -> False, Frame -> True, FrameLabel -> {"\!\(\*SubscriptBox[\(log\), \(e\)]\)(Claim Size)", "Cummulative Probability Density"}, ImageSize -> 400, BaseStyle -> {FontSize -> 14}, AspectRatio -> 1]]
Out[16]=

Marshall Bradley, "Danish Fire Insurance Claims" from the Wolfram Data Repository (2022)  

Data Resource History

Source Metadata

Publisher Information