extRemes (version 2.2)

parcov.fevd: EVD Parameter Covariance

Description

Try to calculate the parameter covariance for an extreme value distribution (EVD) fitted using MLE.

Usage

parcov.fevd(x)

Value

An np by np matrix is returned where np is the number of parameters in the model.

Arguments

x

A list object of class “fevd” as returned by fevd.

Author

Eric Gilleland

Details

Makes possibly two calls to optimHess in an effort to find the parameter covariance matrix for fitted EVDs where MLE is used. The first attempt uses the actual gradient of the negative log-likelihood. If this fails, or the Hessian matrix cannot be inverted, or there are any negative values along the diagonal in the inverted Hessian, then a second attempt is made using finite differences. See Coles (2001) sec. 2.6.4 for more details.

References

Coles, S. (2001) An introduction to statistical modeling of extreme values, London, U.K.: Springer-Verlag, 208 pp.

See Also

fevd, summary.fevd, print.fevd

Examples

z <- revd(100, loc=20, scale=0.5, shape=-0.2)
fit <- fevd(z)
fit

parcov.fevd(fit)