Data Dependent Prior; Gaussian Model

DESCRIPTION:

Computes data dependent (ridge) prior for Gaussian models.

USAGE:

dataDepPrior.preGauss(object, scale, df = 1) 

REQUIRED ARGUMENTS:

object
an object of class "preGauss".
scale
the scale parameter in the inverted-Wishart prior. This is the expected sums of squares and cross-products for the multivariate normal data when there are df degrees of freedom. Any scale matrix of the correct dimensionality for your problem is acceptable. Alternatively, if a scalar value is input for scale, then the final scale matrix is taken to be the diagonal matrix containing the scalar value times the variances of the nonmissing data. (i.e. For each variable, missing values are deleted, and variances calculated on just the observed data.) The usual biased maximum likelihood estimates are computed, see routine colVars.

OPTIONAL ARGUMENTS:

df
the "degrees of freedom" in the scale parameter. This is the number of degrees of freedom used in the inverted Wishart prior distribution for the covariance matrix parameters. The larger the degrees of freedom, the stronger the belief in the prior distribution. Any value greater than -1 is acceptable.

VALUE:

an object of class "priorGauss" containing the expanded scale matrix and other hyperparameters. The components in this class are described in the help file for .

REFERENCES:

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

SEE ALSO:

, .

EXAMPLES:

dataDepPrior.preGauss(object = preGauss(marijuana), scale = 1.4, df = 2)