mixtureCountEvidence

fun ReportBuilder.mixtureCountEvidence(results: MixtureModelingResults, title: String = "Evidence about the number of components", includePlot: Boolean = true)(source)

The evidence for and against each candidate count, with the criteria side by side.

The margin is labelled as the criterion's firmness, not as reliability, because measurement refutes the reading a reader will otherwise take. Across the designed experiment the criterion is confidently wrong when it is wrong — the median gap to the truth is around nineteen — so the size of a margin is not evidence that the choice behind it is sound. What does carry information is whether criteria of different penalty strength agree.

Parameters

results

the modeling results

title

the section title

includePlot

whether to draw the gain-against-penalty figure