plausible
The component counts this sample does not rule out.
Runs testComponentCount at each count considered and collects the ones that survived. See PlausibleComponentCounts for why counts that could not be tested stay out of the set rather than being counted as survivors.
The cost is the whole range's. One test is already about two hundred refits, and this is one per count, so a five-count range is roughly a thousand. Narrow the range when only a particular reading of the data is in question.
Streams are taken one per count so the tests are independent. A streamNum of zero, the convention for "next available", gives each count the next stream from the provider, which is reproducible for a given provider and call order; any other value is used as the first of a consecutive block, one per count considered.
Parameters
the counts to test
how many samples to draw per count
the level at which each verdict is read
a factory for the component fitter
a factory for the refiner
a factory for the family selector
the first stream to draw with, or zero for next-available
the stream provider