gridsearch_burninthining_multiple {ERPM}

Grid - search burnin thining multiple

Description

Function that simulates the Markov chain for a given model and several sets of transitions (the neighborhoods), for multiple partitions. For each neighborhood, it calculates the autocorrelation of statistics for different thinings and the average statistics for different burn-ins. Then the best neighborhood can be selected along with good values for burn-in and thining

Usage

gridsearch_burninthining_multiple(
  partitions,
  presence.tables,
  theta,
  nodes,
  effects,
  objects,
  num.steps,
  neighborhoods,
  numgroups.allowed,
  numgroups.simulated,
  sizes.allowed,
  sizes.simulated,
  max.thining,
  parallel = FALSE,
  cpus = 1
)

Arguments

partitions

Observed partitions

presence.tables

Presence of nodes

theta

Initial model parameters

nodes

Node set (data frame)

effects

Effects/sufficient statistics (list with a vector "names", and a vector "objects")

objects

Objects used for statistics calculation (list with a vector "name", and a vector "object")

num.steps

Number of samples wanted

neighborhoods

List of probability vectors (proba actors swap, proba merge/division, proba single actor move)

numgroups.allowed

vector containing the number of groups allowed in the partition (now, it only works with vectors like num_min:num_max)

numgroups.simulated

vector containing the number of groups simulated

sizes.allowed

Vector of group sizes allowed in sampling (now, it only works for vectors like size_min:size_max)

sizes.simulated

Vector of group sizes allowed in the Markov chain but not necessraily sampled (now, it only works for vectors like size_min:size_max)

max.thining

Where to stop adding thining

parallel

False, to run different neighborhoods in parallel

cpus

Equal to 1

Value

list