fitlgc {lgc}R Documentation

Fit Log Gaussian (Cox) model using formula interface. Can handle both count observations and/or direct oberservations of the Log Gaussian process. The latent field is integrated out using the Laplace approximation. Fixed effects are estimated using a REML criterion.

Description

The formula interface helps building the often very large Gaussian Markov Random Fields. These are represented by sparse matrices.

Usage

  fitlgc(ranef, fixef, data, delta = 1e-04,
    parscale = rep(0.01, length(C@parameters)),
    control = .lgc_control(), update.args = NULL,
    nugef = NULL, etafix = NULL, permfun = NULL,
    scale = NULL, denseFixef = TRUE, betafix = NULL,
    nonlin = NULL, parameters = NULL, ignore.proj = FALSE,
    optimize = TRUE, ...)

Arguments

ranef

Passed to covStructFormula.

fixef

Passed to covStructFormula.

data

Passed to covStructFormula.

delta

Passed to covStructFormula.

parscale

Passed to optim.

control

Object returned by .lgc_control().

update.args

Passed to update method of covStructs.

nugef

Nugget effect formula.

etafix

Formula giving a variable in the dataset with values in the random field to fix. Used to specify direct Log-Gaussian observations. NAs are ignored so that only parts of the field have to be observed.

permfun
scale
betafix

Named vector of fixed effects to hold fixed.

Value

Fitted object of class fitlgc.


[Package lgc version 1.4 Index]