daGauss.default(object, subset, prior = <<see below>>,
start = <<see below>>, control = daGauss.control())
daGauss.preGauss(object, prior = <<see below>>,
start = <<see below>>, control = daGauss.control())
daGauss.missmodel(object, prior = <<see below>>,
start = <<see below>>, control = daGauss.control())
daGauss.default: data frame or matrix containing the raw data.
When a data frame is
input, the model is applied to all numeric variables. When a matrix
is input, all variables are used.
daGauss.preGauss:
an object of class
"preGauss" (produced by the
preGauss
function).
daGauss.missmodel:
an object of class
"missmodel" containing the results of a previous
log-linear analysis. Any of the functions
mdGauss,
completeGauss,
emGauss, or
daGauss may be used to produce the
missmodel object.
object is a data frame,
this expression may use variables in the data frame.
"priorGauss".
"noninformative"
and
"ridge" (for the default ridge prior).
String matching is used, so the characters
"n" or
"r" are
sufficient.
prior="ml" is not allowed for data augmentation, which
samples from the posterior distribution, and therefore requires
a
prior (when
prior="ml", no
prior
is specified and maximum likelihood estimates are produced).
"priorGauss" object is created by routine
priorGauss.
"missmodel" object is input, any value specified in a previous call
has priority over the default value (but not over any currently
specified value).
Gauss functions are the mean and variance--covariance matrix of a multivariate
normal distribution. Thus,
start may be a list with vector
component
mu giving the mean and matrix component
sigma giving
the variance-covariance matrix. Alternatively, a class
"Gauss"
object created as the
paramIter component of the class
"missmodel"
object may be used. Routines
mdGauss,
daGauss, and
emGauss may be
used to create an appropriate
"missmodel" object.
object is
a class
"missmodel" object. In this case argument
start defaults
to the final estimates in the input
"missmodel" object.
daGauss.missmodel: if not given, argument
control defaults to
the control parameters specified in the
call statement of the input
"missmodel" object, but only if these are of the correct class. If
these are not given (or are not of the correct class), then the
argument
control defaults to
daGauss.control values.
"missmodel" is returned; see
for details.
.Random.seed
if it does not already
exist; otherwise they update its value.
See the help file for for additional details.
# Data augmentation: default startings values; save iterates 101 to 1100
daGauss.default(object = cholesterol,
control = list(save = 101:500))
cholesterol.pre <- preGauss(data = cholesterol)
daGauss.preGauss(cholesterol.pre,
control = list(save = 101:500))
# Use as starting values the final estimates given in cholesterol.em
cholesterol.em <- emGauss(object = cholesterol)
daGauss.missmodel(object = cholesterol.em,
control = list(save = 101:500))