glmmControl.defaults: glmm control default values

In MarioniLab/miloR: Differential neighbourhood abundance testing on a graph

glmm control default values

Description

This will give the default values for the GLMM solver

Usage

glmmControl.defaults(...)

Arguments

...	see fitGLMM for details

Details

The default values for the parameter estimation convergence is 1e-6, and the maximum number of iterations is 100. In practise if the solver converges it generally does so fairly quickly on moderately well conditioned problems. The default solver is Fisher scoring, but this will switch (with a warning produced) to the NNLS Haseman-Elston solver if negative variance estimates are found.

Value

list containing the default values GLMM solver. This can be saved in the user environment and then passed to testNhoods directly to modify the convergence criteria of the solver that is used.

theta.tol:

numeric scalar that sets the convergence threshold for the parameter inference - this is applied globally to fixed and random effect parameters, and to the variance estimates.

max.iter:

numeric scalar that sets the maximum number of iterations that the NB-GLMM will run for.

solver:

character scalar that sets the solver to use. Valid values are Fisher, HE or HE-NNLS. See fitGLMM for details.

Author(s)

Mike Morgan

Examples

mmcontrol <- glmmControl.defaults()
mmcontrol
mmcontrol$solver <- "HE-NNLS"
mmcontrol

MarioniLab/miloR documentation built on Oct. 18, 2024, 6:04 p.m.