logpost.Loglin(object, theta = <<see below>>, prior = <<see below>>)
"missmodel" whose
paramIter component is of
class
"Loglin". I.e., an object resulting from the use of the
mdLoglin,
daLoglin, or
emLoglin functions. Alternatively, a class
"preLoglin" object.
paramIter component
of a
missmodel object. If a
class
"preLoglin" object is input, the cell probabilities must be
specified.
"Loglin" object with multiple rows; then calculations
are performed for each row, and a vector is returned.
"priorLoglin", or an array of
hyperparameters.
"ml" (maximum likelihood) or
"noninformative".
String matching is used,
so the characters
"m" or
"n"
are sufficient. The values
of the hyperparameters changes with the algorithm (see
for details). E.g.
"noninformative" means a common value of 1 for
EM, and a common value of 0.5 for DA.
"priorLoglin" object is created by routine
priorLoglin.
start for the order to use in specifying a vector of
hyperparameters. If a single numeric value is input, its value is
replicated for all cells in the table.
The hyperparameters for a data dependent prior (following an
independence model) can be generated using routine
dataDepPrior.
See
for details.
"noninformative". When a class
"missmodel"
object is input, any value specified in a previous call has priority
over the default value (but not over any currently specified value).
NA) when a vector of
hyperparameters is input as argument
prior.
theta.
This is the log-likelihood or log-posterior density that ignores the missing-data mechanism.
fit <- emLoglin(object = crime, margins = ~Visit.1+Visit.2) logpost.Loglin(fit)