Generate a list of settings for Metropolis-Hastings updates and adaptation via the robust adaptive Metropolis algorithm (Vilhola, 2012).

Generate a list of settings for Metropolis-Hastings updates and adaptation via the robust adaptive Metropolis algorithm (Vilhola, 2012).

mvnmh_control(
  n_updates = 1,
  target_acceptance = 0.234,
  max_scaling = Inf,
  scale_constant = 1,
  scale_cooling = 2/3,
  step_size = 1,
  stop_adaptation = 0,
  adaptation_offset = 0,
  nugget = 1e-05,
  nugget_cooling = 2/3,
  nugget_step_size = 1
)

Arguments

n_updates

number of updates per iteration.

target_acceptance

target acceptance rate, defaults to 0.234

max_scaling

maximum global scaling factor, defaults to Inf.

scale_constant

constant multiple of the adaptations determined by scale_cooling.

scale_cooling

rate at which to cool the adaptation, defaults to 0.5. Adaptation contributions are governed by a harmonic sequence: scale_constant/(iteration/step_size+1)^scale_cooling. The plot_adaptations function may be used to plot the adaptation factors.

step_size

adaptation increment for each iteration, defaults to 1.

stop_adaptation

iteration at which adaptation should be terminated, defaults to 0 for no adaptation.

adaptation_offset

iteration offset

nugget

nugget for proposal covariance

nugget_cooling

rate at which to cool the nugget, defaults to 0.9. Adaptation contributions are governed by a harmonic sequence: scale_constant/(iteration/step_size+1)^scale_cooling. The plot_adaptations function may be used to plot the adaptation factors.

nugget_step_size

increment for each iteration, defaults to 100/number of iterations.

Value

list with control settings for Metropolis-Hastings updates