Data Augmentation for Log-Linear Models

DESCRIPTION:

Internal computation routine for data augmentation for log-linear models. This routine is not normally called directly by users.

USAGE:

daLoglin.compute(object, nmarg, prior, start, 
    control = daLoglin.control(), saturated.model) 

REQUIRED ARGUMENTS:

object
a class "preLoglin" object containing the table to be analyzed.
nmarg
a vector containing the indices of the margins to be fit. A margin is described by the factors not summed over, and margins are separated by zeros. Thus c(1,2,0,2,3,0,1,3) would indicate fitting the (1,2), (2,3), and (1,3) margins in a three-way table, i.e., the model of no three-way interaction.
prior
specifies Dirichlet prior hyperparameters. Supply either a character string, or an object of class "priorLoglin", or an array of hyperparameters.

Valid character strings are "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.

A class "priorLoglin" object is created by routine priorLoglin. This routine allows you to easily specify data dependent priors.

See argument 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 default value is "noninformative".

Structural zeros must be coded as missing ( NA) when a vector of hyperparameters is input as argument prior.
start
starting values of the parameters. The parameters estimated by mdLoglin are the cell probabilities. Thus, start is a vector with length equal to the total length of the table containing a probability estimate for each cell in the table. Starting values for cells that are structural zeros in the table should be zero. The default starting values are all equal to one divided by the number of cells in the table. Suppose that the table is defined by the variables X1, X2, and X3. Then the cells in the table are ordered such that the index for variable X1 varies fastest, the index for variable X2 varies next fastest, etc.
control
a list of parameters used to control the algorithm; see for details.
saturated.model
if T, the model is saturated (no further parameters are of use in estimating the cell probabilities). Otherwise, the model is not saturated.

VALUE:

an object of class "missmodel" is returned; see for details.

SIDE EFFECTS:

The function daLoglin.compute creates the data set .Random.seed if it does not already exist, otherwise its value is updated.

DETAILS:

See the help file for for additional details. This function is not normally called directly by users.

SEE ALSO:

, , , , , , , .