Jaatha.refinedSearch: Iterative search for the maximum composite likelihood parameters

Description

This function searches for the parameter combination with the highest composite likelihood. Therefore it iteratively searches the area (=block) around the last found value for a new maximum using simulations and a generalized linear model.

Usage

Jaatha.refinedSearch(jaatha, best.start.pos = 2,
  sim = length(getParNames(jaatha)) * 25, sim.final = min(sim, 100),
  epsilon = NULL, half.block.size = 0.025, weight = NULL,
  max.steps = 200, rerun = FALSE)

Arguments

jaatha

The Jaatha settings (create with Jaatha.initialize)

best.start.pos

This is the number of best starting positions found in the initial search that we will use. Jaatha runs a separate search starting from each of this points.

sim

The number of simulations that are performed in each step

sim.final

The number of simulations that are performed after the search to estimate the composite log likelihood. If not specified, the value of sim will be used

epsilon

Obsolete. Has no effect anymore and will be remove on next major release.

half.block.size

The size of the new block that is created around a new maximum.

weight

Obsolete. Has no effect anymore and will be remove on next major release.

max.steps

The search will stop at this number of steps if not stopped before (see epsilon).

rerun

You can repeat a previously done refined search in Jaatha. Do do so, just call the refined search function with the jaatha object result of the first refined search and set rerun to 'TRUE'.

Value

An Jaatha object. The found values are written to the slot likelihood.table.