Parameters in a Log-linear Model Prior.

DESCRIPTION:

Computes the parameters in a log-linear model prior. This routine is not normally called directly by users.

USAGE:

check.Loglin.prior(prior, algorithm = "em", object) 

REQUIRED ARGUMENTS:

prior
an object of class "priorLoglin" containing the type of the prior and other information, see . The prior type may be one of "ml", "noninformative", or "data.dependent".

OPTIONAL ARGUMENTS:

algorithm
character, either "em" for the EM algorithm, or "da" for data augmentation.
object
an object of class "preLoglin" used for generating data dependent priors.

VALUE:

for algorithm == "da", if the prior type is "ml", check.Loglin.prior assigns a prior of 0 to each cell in the table, while if prior type is "noninformative" then a prior of 0.5 is assigned to each cell. For algorithm == "em", check.Loglin.prior assigns a prior of 1 to each cell in the table with both the maximum likelihood and the "noninformative" prior. When a data dependent prior is used, check.Loglin.prior computes the marginal probabilities for the observed data using the input "preLoglin" object. These marginal probabilities are used to compute the cell probabilities. The data dependent prior is the product of the nobs value found in the "priorLoglin" object, and these cell probabilities.

SEE ALSO:

, .

EXAMPLES:

prior <- check.Loglin.prior(priorLoglin("noninformative")) 
priorObject <- priorLoglin("data.dependent", nPriorObs = 5) 
object <- preLoglin(crime, frequency = count) 
prior <- check.Loglin.prior(priorObject, object = object)