Mode of Log-posterior Distribution

DESCRIPTION:

Computes the mode of the log-posterior distribution defined by a multivariate normal distribution and an normal inverse Wishart prior. This routine is not normally called directly by users.

USAGE:

logpost.Gauss.compute(object, theta, prior, rcmin) 

REQUIRED ARGUMENTS:

object
an object of class "preGauss".
theta
the model parameters. These are usually obtained as the final estimates (the last row) in the paramIter component of a missmodel object (see )

OPTIONAL ARGUMENTS:

prior
an object of class "priorGauss" giving the hyperparameters of the prior distribution. Routine priorGauss is used to create the class "priorGauss" object. Alternatively, the character strings "ml" (for no prior, i.e., maximum likelihood estimation), "noninformative" (for a noninformative prior), or "ridge" (for the default ridge prior) may be used. Pattern matching means that only the first character in the string is required. See for details.
rcmin
parameter used to determine when the variance-covariance matrix is singular. Typically, sqrt(.Machine$double.eps) is acceptable.

VALUE:

the computed mode of the log-posterior distribution.

DETAILS:

This routine is not normally called directly by users.

REFERENCES:

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

SEE ALSO:

, , , .