cov.family
contains
information required to optimize a profile likelihood in a spatial
regression analysis, and is required input for
slm
.
No functions for constructing objects of class
"cov.family"
exist. Objects
of the class have already been created for the standard covariance structures,
CAR
,
SAR
, and
MA
. If you choose to construct a new
"cov.family"
object, the easiest way to proceed is to copy an
existing
"cov.family"
object, and modify the new object.
The class
"cov.family"
has no associated methods.
Class
"cov.family"
does not inherit from any other class.
The
"cov.family"
object is implemented as a list with the
components given below.
det.fun(parameters, spatial.arglist, initial)
parameters
- a vector containing the current values of the covariance matrix model
parameters. A profile likelihood for the covariance matrix model parameters is optimized
given optimal values for the parameters in the linear model.
spatial.arglist
- a list of additional arguments to be used in the
spatial model of the covariance matrix. See routine
slm
.
initial
- a list containing the return value from the
initialize
function described below.
y
in the linear equation
x = S y
, where
S
is a square symmetric covariance matrix, and
x
is a known
matrix. The solution of this equation defines the spatial covariance
matrix
S
.
solve.fun(parameters, x, spatial.arglist, initial)
parameters
- a vector containing the current values of the spatial parameters required by
the model for the covariance matrix (see above).
x
- a matrix containing the right hand sides.
spatial.arglist
- a list of additional arguments to be used in
defining the spatial model for the covariance matrix. See routine
slm
.
initial
- a list containing the return value from the
initialize
function described below. These may be used as additional variables
when computing the spatial covariance matrix.
x
to the linear equations
y = S x
, where
S
is the spatial covariance matrix specified by this family of
covariance matrices.
initial.parameters(spatial.arglist)
spatial.arglist
- a list of additional arguments to be used in the
spatial model of the covariance matrix. See routine
slm
.
initialize(spatial.arglist, subset)
spatial.arglist
- a list of additional arguments to be used in the
spatial model of the covariance matrix. See routine
slm
.
subset
- a logical vector (which is replicated to have length equal
to the number of observations), a numeric vector, or a character
vector of the row names. This vector is used to indicate
which observation numbers are to be included in the analysis. All
observations are included by default. See routine
slm
.
det.fun
and the
solve.fun
to
simplify and speed the computations. For many covariance models, this
is the eigenvalues of the symmetric spatial neighbor matrix.
cov.family
to allow cleanup operations, should these be necessary.
finalize(spatial.arglist, initial)
spatial.arglist
- a list of additional arguments to be used in the
spatial model of the covariance matrix. See routine
slm
.
initial
- a list containing the values returned by the
cov.family$initialize
function. This list depends upon the
cov.family
.
residual.fun(x, y, beta, parameters, spatial.arglist)
x
- the independent variable(s) (if any).
y
- the dependent variable.
spatial.arglist
- a list of additional arguments to be used in the
spatial model of the covariance matrix. See routine
slm
.
The six functions,
det.fun
,
solve.fun
,
residual.fun
,
initial.parameters
,
initialize
, and
finalize
are used by the
spatial regression routine to define a spatial regression. Currently
three models for a spatial regression can be fit: CAR, SAR, and MA.