SliceMCMC

Purpose

SliceMCMC uses directional slice-style proposals.

Full Sampling Section Keys

• Sampling.Method (required): must be SliceMCMC.
• Sampling.Variables (required):
• name, description, distribution.type, distribution.parameters
• runtime-safe parameter sets: Flat(min,max), Log(min,max), Normal(mean,stddev), Log-Normal(mean,stddev), Logit(location,scale)
• Sampling.LogLikelihood (required): array of {name, expression}
• Sampling.selection (optional, string)
• Sampling.Bounds:
• base keys: num_chains, num_iters, proposal_scale
• slice keys:
  • slice_mode (optional, string, default random_direction)
  • slice_width (optional, number, default 0.2)
  • slice_max_steps_out (optional, integer, default 16)
  • slice_max_shrink (optional, integer, default 32)
  • slice (optional object alias): mode, width, max_steps_out, max_shrink

Full Skeleton

Sampling:
  Method: "SliceMCMC"
  Variables:
    - name: x
      description: variable x
      distribution:
        type: Flat
        parameters:
          min: -5
          max: 5
  LogLikelihood:
    - name: L_x
      expression: "-0.5*(x/1.0)^2"
  Bounds:
    num_chains: 6
    num_iters: 9000
    proposal_scale: [0.08, 0.08, 0.08, 0.08, 0.08, 0.08]
    slice_mode: random_direction
    slice_width: 0.25
    slice_max_steps_out: 20
    slice_max_shrink: 40
    slice:
      mode: random_direction
      width: 0.25
      max_steps_out: 20
      max_shrink: 40

Example

Sampling:
  Method: "SliceMCMC"
  Variables:
    - name: x
      description: variable x
      distribution:
        type: Flat
        parameters:
          min: -5
          max: 5
  LogLikelihood:
    - name: L_x
      expression: "-0.5*(x/1.0)^2"
  Bounds:
    num_chains: 6
    num_iters: 9000
    proposal_scale: [0.08, 0.08, 0.08, 0.08, 0.08, 0.08]
    slice_mode: random_direction
    slice_width: 0.25
    slice_max_steps_out: 20
    slice_max_shrink: 40