sampleChains

sampleChains(modelFactory, options?): Promise<{ acceptanceRates: number[]; byChain: any; chains: any[]; parallel: boolean; seeds: number[]; trace: any; }>

Defined in: parallel.js:260

Sample several MCMC chains in parallel, one worker per chain.

Accepts a SELF-CONTAINED model factory (data, mc) => Model — see the module doc for the contract — and returns the per-chain fits plus pooled and by-chain traces ready for gelmanRubin / effectiveSampleSize.

Falls back to running the chains sequentially in-process (identical numbers, same per-chain seeds) when the runtime has no workers or when parallel: false is passed.

Parameters

modelFactory

(data, mc) => Model

Builds the model from data and the mc toolkit; must not reference outer variables.

options?

chains?

number

Number of chains

data?

any

Structured-clonable data for the factory

inits?

any

Per-chain initial values (array, one per chain — over-dispersed starts recommended), or a single init object used for every chain. Required.

nSamples?

number

Draws per chain

nWarmup?

number

Warmup iterations per chain

parallel?

boolean

Force the sequential in-process path with false (same results, no workers)

sampler?

string

‘nuts’ | ‘hmc’ | ‘metropolis’

samplerOptions?

any

Constructor options for the sampler (e.g. {stepSize, maxTreeDepth, targetAcceptance})

seed?

number

Base seed; chain c uses a seed derived from it, so runs are reproducible and chains are independent

thin?

number

Thinning interval

Returns

Promise<{ acceptanceRates: number[]; byChain: any; chains: any[]; parallel: boolean; seeds: number[]; trace: any; }>