KSLCore/ksl.utilities.distributions.fitting.mixture/MixtureDataDescription

MixtureDataDescription

class MixtureDataDescription(val numObservations: Int, val numDistinctValues: Int, val statistics: StatisticIfc, val maximumFeasibleComponents: Int, val modality: ModalityAssessment)(source)

Everything known about a sample before a mixture is fitted to it.

The order matters. An analyst is best served by learning that their sample cannot support six components before being shown a fit at six components, and by learning that the humps they counted move with the bin width before being asked to defend the count. Both facts are cheap; neither is available from the fit.

Adequacy is deliberately absent. The plan expected the data-adequacy verdict to join this object, and it cannot: adequacy is governed by how far the fitted mixture sits from the nearest mixture with one component fewer, so it requires a fit that by definition does not exist yet. It lives on the results instead. What belongs here is the structural ceiling — the largest number of components this sample can support at all — which is available before fitting and answers the same question in the cases where the answer is emphatic.