fit_p2_IS: Improve the fit of p2 models

In rugbel/p2model: Maximum Likelihood Estimation of p2 models

Description

Estimates a p2 model using parameter dependent Laplace Importance Sampling.

Usage

fit_p2_IS(objfit, M = 5000, ncores = 1, init = NULL,
  init.hess = TRUE, seed = NULL, trace = 0, hessian = 1,
  grtol = 1e-06, xtol = 1e-12)

Arguments

objfit	An object produced by a call to fit_p2.
M	number of importance sampling simulations. Default is M=5000.
ncores	Number of cores to employ.
init	Starting point for the model parameter value. If it is NULL, the theta slot of objfit is used.
init.hess	Should the hessian matrix from fit_p2 be used as the initial Hessian by the ucminf optimiser? Default is TRUE.
seed	Random seed for the importance sampler.
trace, hessian, grtol, xtol	Arguments for the ucminf optimiser. For their meaning, see the help file for that function.

Details

The function can be called only after fitting a model by fit_p2, so that all the model settings are inherited from those employed in the call tofit_p2. The function can take advantage of multiple cores for parallel evaluation of the samples simulated by the importance sampler. The optimiser used is ucminf, which can exploit the estimated Hessian from fit_p2 in a Quasi-Newton optimisation.

Value

The returned value is an object of class "p2", a list containing the same components listed for fit_p2.

Author(s)

Ruggero Bellio

See Also

fit_p2, ucminf

Examples

## Not run: 
mod <- fit_p2(Y, Xn, Xn, XvD, XvC)
nc <- parallel::detectCores()
modIS <- fit_p2_IS(mod, ncores=nc)
## End(Not run)

rugbel/p2model documentation built on March 8, 2021, 8:16 p.m.