- model
a R function implementing the model to be simulated. It
must take as arguments a vector of model parameter values and it must return
a vector of summary statistics. When using the option use_seed=TRUE,
model must take as arguments a vector containing a seed value and the
model parameter values. A tutorial is provided in the package's vignette to
dynamically link a binary code to a R function. Users may
alternatively wish to wrap their binary executables using the provided
functions binary_model and binary_model_cluster.
The use of these functions is associated with slightly different constraints
on the design of the binary code (see binary_model and
binary_model_cluster).
- prior
a list of prior information. Each element of the list
corresponds to a model parameter. The list element must be a vector whose
first argument determines the type of prior distribution: possible values
are "unif" for a uniform distribution on a segment, "normal"
for a normal distribution, "lognormal" for a lognormal distribution
or "exponential" for an exponential distribution. The following
arguments of the list elements contain the characteritiscs of the prior
distribution chosen: for "unif", two numbers must be given: the
minimum and maximum values of the uniform distribution; for "normal",
two numbers must be given: the mean and standard deviation of the normal
distribution; for "lognormal", two numbers must be given: the mean
and standard deviation on the log scale of the lognormal distribution; for
"exponential", one number must be given: the rate of the exponential
distribution. User-defined prior distributions can also be provided. See the
vignette for additional information on this topic.
- nb_design_pts
a positive integer equal to the desired number of
simulations of the model used to build the emulator.
- nb_simul
a positive integer equal to the desired number of
simulations of the emulator.
- 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. If not provided, no constraint
will be applied.
- summary_stat_target
a vector containing the targeted (observed)
summary statistics. If not provided, ABC_rejection only launches the
simulations and outputs the simulation results.
- emulator_span
a positive number, the number of design points selected
for the local regression. 50 by default.
- tol
tolerance, a strictly positive number (between 0 and 1)
indicating the proportion of simulations retained nearest the targeted
summary statistics.
- use_seed
logical. If FALSE (default), ABC_rejection
provides as input to the function model a vector containing the model
parameters used for the simulation. If TRUE, ABC_rejection
provides as input to the function model a vector containing an
integer seed value and the model parameters used for the simulation. In
this last case, the seed value should be used by model to initialize
its pseudo-random number generators (if model is stochastic).
- 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.
- n_cluster
a positive integer. If larger than 1 (the default value),
ABC_rejection will launch model simulations in parallel on
n_cluster cores of the computer.
- verbose
logical. FALSE by default. If TRUE,
ABC_rejection writes in the current directory intermediary results at
the end of each step of the algorithm in the file "output". These outputs
have a matrix format, in wich each raw is a different simulation, the first
columns are the parameters used for this simulation, and the last columns
are the summary statistics of this simulation.
- progress_bar
logical, FALSE by default. If TRUE,
ABC_rejection will output a bar of progression with the estimated
remaining computing time. Option not available with multiple cores.