glcSplitCriterionLevelWtdBIC: Gaussian RPMM Split Criterion: Level-Weighted BIC
Description
Split criterion function: use a level-weighted version of BIC to determine split; there is an additional penalty incorporated for deep recursion.
Usage
glcSplitCriterionLevelWtdBIC(llike1, llike2, weight, ww, J, level)
Arguments
llike1
one-class likelihood.
llike2
two-class likelihood.
weight
weights from RPMM node.
Value
- bic1
- One-class BIC, with additional penalty for deeper levels
- bic2
- Two-class BIC, with additional penalty for deeper levels
- split
TRUE=split the node, FALSE=do not split the node.
Details
This is a function of the form “glcSplitCriterion...”, which is required to return a list
with at least a boolean value split, along with supporting information.
See glcTree for example of using “glcSplitCriterion...” to control split.