Data Dependent Prior; Log-linear Model

DESCRIPTION:

Computes data dependent prior for log-linear models.

USAGE:

dataDepPrior.preLoglin(object, nPriorObs, algorithm = "da") 

REQUIRED ARGUMENTS:

object
an object of class "preLoglin"
nPriorObs
the number of prior observations. Intuitively, this is the number of observations added by the prior probabilities.
algorithm
character, either "em" for the EM algorithm, or "da" for data augmentation.

VALUE:

a vector containing the Dirichlet hyperparameters for each cell.

DETAILS:

The prior probabilities are computed assuming all variables are independent. For each variable, the marginal probability is computed using the observed data for that variable.

REFERENCES:

Schafer, J. L. (1997), Analysis of Incomplete Multivariate Data, Chapman & Hall, London.

SEE ALSO:

, .

EXAMPLES:

dataDepPrior.preLoglin(object = preLoglin(crime), nPriorObs = 5, 
                              algorithm = "da")