profile.etasclass: profile method for etasclass objects (ETAS model) (To be checked)

Description

profile method for etasclass objects (ETAS model).

Usage

# S3 method for etasclass
profile(fitted,iprofile		=1,
				 nprofile	=7,
				 kprofile	=3,
				 profile.approx	=FALSE,...)

Value

Returns a list:

params.vec: vector of values of the parameter iprofile used to evaluate the profile likelihood.
logl.vec: vector of likelihoods corresponding to the values of params.vec

plot method is defined to represent profile likelihood (in scale -2log(LR)), using a spline interpolation through grid points, with superimposition of approximate confidence intervals.

Arguments

fitted: An object of the class etasclass

iprofile: An integer in the range 1-7. Profile likelihood will be computed with respect to the parameter of index iprofile. The order of parametrs is: mu, k0, c, p, a, gamma, d, q. In this version 2.0.0 profile only for etas parameters, not for covariates parameters. It will added future versions
nprofile: Number of values of params[iprofile] for which profile likelihood must be computed. Default value= 7.
kprofile: Maximum absolute standardized value for params[iprofile]. Profile likelihood will be computed in the standardized range [-kprofile, kprofile]. Default value= 3.
profile.approx: if TRUE, then a conditional-likelihood approach is used as a first value for each maximization step in profile likelihood computation. Default value= FALSE.
...: other arguments.

Author

Marcello Chiodi, Giada Adelfio

Details

Profile likelihood for the iprofile-th parameter of the ETAS model for earthquake seismicity, estimated with etasclass; the order of parameters is: mu,k0,c,p,gamma,d,q and betacov.

A plot method is defined for profile.etasclass objects. A number of grid points nprofile of 7 (the default) usually is enough to have a good interpolation of the profile likelihood. The profile is computed using the final estimation of the background seismicity used to obtain the object etas of class etasclass and regardless to the method used. The computing time (for each of the nprofile values) is generally less than a single execution of etasclass without clustering, because only ML estimation is performed. Parameters not estimated in etas (with params.ind[i]=FALSE) will remain fixed do the value params.fix[i].