Skip to contents

Bayesian MoE (bmoe) Arguments

bmoe-args.Rd

Helper functions to assist in supplying arguments to bmoe().

Usage

bmoe_prior(k, ...)

bmoe_jags_n(...)

bmoe_inits(seed_base, .RNG.name = "base::Wichmann-Hill", ...)

Arguments

k

integer. Assumed number of components.

...

Overwrite default values.

seed_base

numeric. Seed are set equal to seed_base multiplied by chain index.

.RNG.name

character. JAGS RNG method.

Prior

Hyper-parameters that must be passed to the prior argument are in bold.

  • k, number of components, assumed known.

  • regr is element-wise IID Normal with 0 mean and regr_prec precision.

  • wt is element-wise IID Normal with 0 mean and wt_prec precision.

  • prec is element-wise IID Gamma with prec_shape and prec_rate.

JAGS Controls

  • n.adapt controls number of discarded samples in adaptation stage.

  • n.update controls number of discarded samples in warm-up stage.

  • n.iter controls how many samples are saved.

  • n.thin controls thinning, where only every \(n^th\) sample is kept.

  • n.chains controls number of chains.

JAGS Initial Values

Optional chain starting values, as described in rjags::jags.model().