OpVaR (version 1.2)

fitMixing: Maximum Likelihood Estimation

Description

Maximum Likelihood Estimation for the dynamic weighted mixture model of Frigessi et al., 2002

Usage

fitMixing(cell, body, tail, method="L-BFGS-B", c_location0=0.75, c_scale0=2)

Arguments

cell

lossdat cell

body

body distribution, either gamma, lnorm or weibull

tail

tail distribution, either gamma, lnorm, weibull or gpd

method

optimization method, default is "L-BFGS-B"

c_location0

empirical quantile of loss severity data used for initialization of Cauchy location parameter in optimization: quantile(cell$Loss,c_location0)

c_scale0

scaling factor for empirical standard deviation used for initialization of Cauchy scale parameter in optimization: sd(cell$Loss)/c_scale0

Value

Returns a sevdist object of type 'mixing' with the given body and tail distributions fitted to the loss data.

Details

Body and tail parameters are initialized by method of moments estimators. Cauchy location is initialized by empirical 70

References

Frigessi et al. A Dynamic Mixture Model for Unsupervised Tail Estimation without Threshold Selection, Extremes 5(3):219-235, 2003

Examples

# NOT RUN {
data(lossdat)
sev=fitMixing(lossdat[[1]],"weibull","gpd")
sev
plot(sev,5000)
# }