Extract Parameters from a missmodel Object

DESCRIPTION:

Extract and label parameters from the paramIter component of a missmodel object.

Function paramIter is generic. Methods currently exist for missmodel objects, and for the following specific classes of paramIter components of missmodel objects: cgm, Gauss and Loglin.

USAGE:

paramIter(object, ...) 
paramIter.missmodel(object, ...) 
paramIter.default(object, save = dim(object)[1]) 
paramIter.cgm(object, save = dim(object)[1], expand = F) 
paramIter.Gauss(object, save = dim(object)[1], expand = F) 
paramIter.Loglin(object, save = dim(object)[1]) 

REQUIRED ARGUMENTS:

object
Either a missmodel object, or a paramIter component of a missmodel object.

OPTIONAL ARGUMENTS:

...
additional arguments to pass to methods for specific classes of paramIter components.
save
a vector giving the rows (iterations) in the parameter matrix to be extracted. The default is to extract the last row.
expand
logical value; if TRUE, then the parameters are expanded into a more readily understood form.

VALUE:

the requested parameters. For a Loglin model, the parameters are organized as an array, where each "row" (the first dimension) is for an iteration and remaining dimensions form a table of estimated cell probabilities for that iteration.

For other models, if expand=FALSE, the value is a matrix with class "Gauss" or "cgm" in which each row contains parameter estimates from one iteration.

If expand=TRUE the value is a list of lists, each inner list containing parameters for one iteration. For a Gauss model each inner list contains a vector of means and a variance-covariance matrix. For a cgm model each inner list contains a matrix of means (one row for each continuous variable, and one column for each cell in the contingency table), a variance-covariance matrix, and a vector of cell probabilities.

If only a single iteration is requested, instead of a list of a single list the single "inner" list is returned; this is subject to change.

SEE ALSO:

, , , , , .

EXAMPLES:

cgmFit <- emCgm(object = stlouis[,-1], margins = ~D1:D2+risk, 
             gauss = ~verbal1+verbal2, design = ~D1+D2+risk, 
             subset = verbal2 > 100 | is.na(verbal2)) 
paramIter(cgmFit) 
gaussFit <- emGauss(cholesterol) 
paramIter(gaussFit) 
estimates <- paramIter(gaussFit, expand=T) 
estimates$mu 
estimates$sigma