gsolv: Estimating function for BR estimation

In rugbel/BRTemplate: Template for BR Estimation for Independent Data Models

Description

Estimating function for BR estimation following Kosmidis and Firth (2010).

Usage

gsolv(para, X, y, ncores, ADobj, dll, datagen, R, seed = NULL,
  observed = TRUE)

Arguments

para	Parameter vector.
X	Model matrix.
y	Observed response.
ncores	Number of cores to use. Recourse to the parallel::detectCores function if you're unsure about them.
ADobj	Object returned by TMB::MakeADFun based on the model specified by the dll file. See the vignette file.
dll	Name of the C++ template representing the model of interest.
datagen	Function that generates a single data set from the model of interest, based on the the specified parameter value and the model matrix.
R	Number of simulated data sets for Monte Carlo computation of expected values. Not used if seed is NULL.
seed	Random seed for Monte Carlo computation. Default is NULL, which switches to the empirical approximation.
observed	Should the observed information used adopted for the direction of the update in the estimation algorithm? If FALSE, the expected Fisher information is used instead. Default is TRUE.

Details

The function is written for usage with the nleqslv package. Both quasi Newton-Raphson and quasi Fisher-scoring update are supported.

Value

The estimating equation at para.

Author(s)

Ruggero Bellio

References

Kosmidis, I. and Firth, D. (2010). A generic algorithm for reducing bias in parametric estimation. Electron. J. Statist., 4, 1097-1112.

Examples

# See the vignette file

rugbel/BRTemplate documentation built on Nov. 12, 2019, 7:41 a.m.