componentsAreSeparable

Whether the recommended mixture's components are ones the data can actually tell apart, or null when the comparison could not be made.

The integrated completed likelihood is the Bayesian criterion plus twice the classification entropy, so the two differ only in whether a component that improves the density but not the classification is worth its cost. Where the entropy-penalised criterion prefers fewer components than the plain one, the components the plain criterion wanted are ones the model can fit but cannot separate: the extra structure improves the density without improving the assignment of observations to components.

Both criteria are already computed, so this is a comparison rather than a computation. It matters for interpretation rather than for the fit: a mixture used as a flexible density does not need separable components, while a mixture read as a description of subpopulations does.

True when the two agree, so components are separable as far as this comparison can tell.

Parameters