Skip to contents

Add an Approximate Bayesian Computation (Monte-Carlo Markov-Chain) simdesign using the Majoram algorithm to a nl object

Source: R/simdesign_helper.R
simdesign_ABCmcmc_Marjoram.Rd

Add an Approximate Bayesian Computation (Monte-Carlo Markov-Chain) simdesign using the Majoram algorithm to a nl object

Usage

simdesign_ABCmcmc_Marjoram(
  nl,
  postpro_function = NULL,
  summary_stat_target,
  prior_test = NULL,
  n_rec,
  n_between_sampling = 10,
  n_cluster = 1,
  use_seed = FALSE,
  dist_weights = NULL,
  n_calibration = 10000,
  tolerance_quantile = 0.01,
  proposal_phi = 1,
  seed_count = 0,
  progress_bar = FALSE,
  nseeds
)

Arguments

nl

nl object with a defined experiment

postpro_function

default is NULL. Allows to provide a function that is called to post-process the output Tibble of the NetLogo simulations. The function must accept the nl object with attached results as input argument. The function must return a one-dimensional vector of output metrics that corresponds in length and order to the specified summary_stat_target.

summary_stat_target

a vector of target values in the same order as the defined metrics of the experiment

prior_test

a string expressing the constraints between model parameters. This expression will be evaluated as a logical expression, you can use all the logical operators including "<", ">", ... Each parameter should be designated with "X1", "X2", ... in the same order as in the prior definition. Set to NULL to disable.

n_rec

Number of samples along the MCMC

n_between_sampling

a positive integer equal to the desired spacing between sampled points along the MCMC.

n_cluster

number of cores to parallelize simulations. Due to the design of the EasyABC parallelization it is currently not possible to use this feature with cores > 1.

use_seed

if TRUE, seeds will be automatically created for each new model run

dist_weights

a vector containing the weights to apply to the distance between the computed and the targeted statistics. These weights can be used to give more importance to a summary statistic for example. The weights will be normalized before applying them. Set to NULL to disable.

n_calibration

a positive integer. This is the number of simulations performed during the calibration step. Default value is 10000.

tolerance_quantile

a positive number between 0 and 1 (strictly). This is the percentage of simulations retained during the calibration step to determine the tolerance threshold to be used during the MCMC. Default value is 0.01.

proposal_phi

a positive number. This is a scaling factor defining the range of MCMC jumps. Default value is 1.

seed_count

a positive integer, the initial seed value provided to the function model (if use_seed=TRUE). This value is incremented by 1 at each call of the function model.

progress_bar

logical, FALSE by default. If TRUE, ABC_mcmc will output a bar of progression with the estimated remaining computing time. Option not available with multiple cores.

nseeds

number of seeds for this simulation design

Value

simdesign S4 class object

Details

This function creates a simdesign S4 class which can be added to a nl object.

Variables in the experiment variable list need to provide a numeric distribution with min, max and a shape of the distribution (qunif, qnorm, qlnorm, qexp)(e.g. list(min=1, max=4, qfun="qunif")).

The function uses the EasyABC package to set up the ABC_mcmc function. For details on the ABC_mcmc function parameters see ?EasyABC::ABC_mcmc Finally, the function reports a simdesign object.

Approximate Bayesian Computation simdesigns can only be executed using the run_nl_dyn function instead of run_nl_all or run_nl_one.

Examples


# To attach a simdesign, a nl object needs to be created first (see ?nl).
# For this example, we load a nl object from test data.

nl <- nl_lhs

# Attach the simdesign to the nl object
nl@simdesign <- simdesign_ABCmcmc_Marjoram(nl = nl,
                                            summary_stat_target = c(100, 80),
                                            n_rec = 100,
                                            n_between_sampling = 10,
                                            n_cluster = 1,
                                            use_seed = FALSE,
                                            n_calibration = 10000,
                                            tolerance_quantile = 0.01,
                                            proposal_phi = 1,
                                            progress_bar = FALSE,
                                            nseeds = 1)
#> Creating ABC Monte-Carlo Markov-Chain simulation design