getLikelihoodProfile: Extract a likelihood profile for the covariance parameters in...

In apear9/SSNdesign: An R Package for Optimal and Adaptive Sampling Designs on Stream Networks

Description

This function computes the likelihood surface in the vicinity of the maximum likelihood estimates of the covariance parameters. Note: this is experimental. This function was initially developed as a workaround to the problem of defining log-normal priors for the covariance parameters when the standard errors extracted by getExpectedSE were manifestly unreasonable.

Usage

getLikelihoodProfile(glmssn, which.vary, grid.parameters,
  parallelism = "none", n.cores = 1, ...)

Arguments

glmssn	An object of class glmssn
which.vary	A vector indicating which of the covariance parameters should be varied. You must vary at least one parameter but you can vary all of them if it suits you.
grid.parameters	An OPTIONAL argument. You can pass a list giving the grid of parameter values for which to evaluate the log-likelihood surface. This can be left missing. In this case, the function attempts to guess the grid range and resolution based on the log-scale confidence-interval on each of the varying parameters.
parallelism	An OPTIONAL argument. This argument indicates what operating system a parallel cluster should be set up for. This option can be one of "none" (for no parallelism), "osx/linux" or "windows". This is not case sensitive. The input should match the user's operating system. Note that this argument is ignored if the next argument n.cores = 1.
n.cores	An OPTIONAL argument. This argument specifies how many CPUs should be assigned to compute the log-likeihoods in parallel. This argument can be any positive integer.
...	Any additional arguments to foreach. Users are NOT recommended to fiddle with this.

Value

A data.frame with one column for each covariance parameter and another column for the log-likelihood.

Author(s)

Jay Ver Hoef and Alan R. Pearse

apear9/SSNdesign documentation built on Feb. 19, 2020, 4:29 a.m.