nadiv (version 2.18.0)

pcc: REML convergence checks

Description

Mainly checks to ensure the variance components in a REML mixed model do not change between the last two iterations more than what is allowed by the tolerance value. See details for extra check on asreml-R models.

Usage

pcc(object, traces = NULL, tol = 0.01, silent = FALSE)

Value

Returns TRUE if all variance parameters change less than the value specified by tol, otherwise returns FALSE. Also see the

details section for other circumstances when FALSE might be returned.

Arguments

object

A list with at least one element named: monitor (see Details)

traces

Optionally, a matrix to substitute instead of the monitor element to object. Each row corresponds to a different variance component in the model and each column is a different iteration of the likelihood calculation (column 1 is the first iterate).

tol

The tolerance level for which to check against all of the changes in variance component parameter estimates

silent

Optional argument to silence the output of helpful (indicating default underlying behavior) messages

Author

Details

Object is intended to be an asreml-R model output. NOTE, The first 3 rows are ignored and thus should not be variance components from the model (e.g., they should be the loglikelihood or degrees of freedom, etc.). Also, the last column is ignored and should not be an iteration of the model (e.g., it indicates the constraint).

The function also checks object to ensure that the output from the asreml-R model does not contain a log-likelihood value of exactly 0.00. An ASReml model can sometimes fail while still returning a monitor object and TRUE value in the converge element of the output. This function will return FALSE if this is the case.

Examples

# Below is the last 3 iterations from the trace from an animal model of 
# tait1 of the warcolak dataset.
# Re-create the output from a basic, univariate animal model in asreml-R
   tracein <- matrix(c(0.6387006, 1, 0.6383099, 1, 0.6383294, 1, 0.6383285, 1),
	nrow = 2, ncol = 4, byrow = FALSE)
   dimnames(tracein) <- list(c("ped(ID)!ped", "R!variance"), c(6, 7, 8, 9))

   pcc(object = NULL, trace = tracein)