findAllMCMCpars {extRemes}

Manipulate MCMC Output from fevd Objects

Description

Manipulates the MCMC sample from an “fevd” object to be in a unified format that can be used in other function calls.

Usage

findAllMCMCpars(x, burn.in = 499, qcov = NULL, ...)

Arguments

x

Object of class “fevd” with component method = “Bayesian”.

burn.in

Burn in period.

qcov

Matrix giving specific covariate values. See 'make.qcov' for more details. If not suplied, original covariates are used.

...

Not used.

Details

This function was first constructed for use by postmode, but might be useful in other areas as well. It evaluates any parameters that vary according to covariates at the values supplied by qcov or else at the covariate values used to obtain the original fit (default). If a model does not contain one or more parameters (e.g., the GP does not have a location component), then a column with these values (set to zero) are returned. That is, a matrix with columns corresponding to location, scale, shape and threshold are returned regardless of the model fit so that subsequent calls to functions like fevd can be made more easily.

This function is intended more as an internal function, but may still be useful to end users.

This function is very similar to findpars, but is only for MCMC samples and returns the entire MCMC sample of parameters. Also, returns a matrix instead of a list.

Value

A matrix is returned whose rows correspond to the MCMC samples (less burn in), and whose columns are “location” (if no location parameter is in the model, this column is still given with all values identical to zero), “scale”, “shape” and “threshold”.

Author(s)

Eric Gilleland

See Also

fevd, findpars, postmode