Log-posterior Density for Multivariate Normal Models

DESCRIPTION:

Calculates the observed-data log-likelihood or log-posterior density for incomplete numeric data modeled by a multivariate normal distribution, with a normal inverse Wishart prior.

USAGE:

logpost.Gauss(object, theta = <<see below>>, prior = <<see below>>) 

REQUIRED ARGUMENTS:

object
an object of class "missmodel" whose paramIter component is of class "Gauss". I.e., an object resulting from the use of the mdGauss, daGauss, or emGauss functions. Alternatively, a class "preGauss" object.

OPTIONAL ARGUMENTS:

theta
the model parameters. If not specified, these are obtained as the final estimates (the last row) in the paramIter component of the input missmodel object. If a class "preGauss" object is input, the model parameters must be specified exactly as in a class "Gauss" object (see ).

This may be a "Gauss" object with multiple rows; then calculations are performed for each row, and a vector is returned.
prior
specifies normal inverted-Wishart prior hyperparameters. Supply either a character string, or an object of class "priorGauss".

Valid character strings are "ml" (maximum likelihood), "noninformative", and "ridge" (for the default ridge prior). String matching is used, so the characters "m", "n", or "r" are sufficient.

A class "priorGauss" object is created by routine priorGauss.

The default value is a noninformative prior. 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).

VALUE:

the value of the observed-data log-likelihood or log-posterior density function evaluated at theta.

DETAILS:

This is the log-likelihood or log-posterior density that ignores the missing-data mechanism.

SEE ALSO:

, , , , , .

EXAMPLES:

fit <- emGauss(cholesterol) 
logpost.Gauss(fit)