STARTlik: Likelihood estimation for starting points

In johannabertl/ApproxML: Approximate maximum likelihood estimation

Description

The function STARTlik draws starting points from a given matrix of intervals (randomly sampled from a uniform distribution or on a grid). For each starting point, the likelihood is estimated from simulations.

Usage

STARTlik(int, sampling, N1, s.obs, simfun, Hfun = bw.nrd0.mult, kernel,
  nk, ...)

LIK(theta, simfun, nk, Hfun, kernel, s.obs, ...)

Arguments

int	matrix of intervals (2 columns, p rows)
sampling	character. One of "random" or "grid"
N1	number of starting points
s.obs	observed summary statistics
simfun	name of the function which is used for the simulation and computes the summary statistics
Hfun	function which computes the bandwidth matrix. Can be one of the functions in bw.R
kernel	kernel function for KDE
nk	number of simulations for the estimation of the likelihood with kernel density estimation
...	further parameters for simfun

Details

For points on a grid, the number of starting points N1 is approximate. For each dimension, round(N1^(1/p)) equally spaced points in the interval are chosen, with p being the number of dimensions.

Note that this could easily be parallelized. This is currently not implemented, since it depends on the system the program is run on.

Functions

• LIK: The function LIK estimates the likelihood at a given point. It is used in SIMlik.

Examples

int = matrix(c(-2, 5, 0, 2), ncol=2, nrow=2, byrow=T)

test_start_grid = STARTrandomlik(int, sampling="grid", N1=20, s.obs = c(2.2, 9), simfun = SIMpoisson_glmm, kernel = robust.unscaled.diagonal, nk=20, x = 1:3)

test_start_random = STARTrandomlik(int, sampling="random", N1=20, s.obs = c(2.2, 9), simfun = SIMpoisson_glmm, kernel = robust.unscaled.diagonal, nk=20, x = 1:3)

johannabertl/ApproxML documentation built on May 22, 2019, 2:19 p.m.