This function uses information from a parameter data frame that stores the
model specifications and aFIMSFrame object that stores the data to
instantiate, i.e., create an instance of a class, the required C++ modules.
Several C++ modules are needed to run a FIMS model and the required modules
will be different for each model type. For example, for a catch-at-age
model one needs to instantiate recruitment, growth, and maturity modules and
at least one fleet and population module.
Arguments
- parameters
A tibble returned from
setup_default_parameters(). It is the primary source of information for what is initialized. That is, if a fleet exists in the data but parameter information for how to specify selectivity for that fleet is not provided, then selectivity will not be initialized for that fleet. The rows of a parameter with more than 1 value are used in order oftiming, thenage, thenlength, not in the order they appear in the tibble. So each row needs its own combination oftiming,age, andlength.- data
An S4 object with the
FIMSFrameclass, which is returned fromFIMSFrame(). Passing the data is required because initialization of the modules requires passing the data and information regarding the uncertainty of that data, i.e., input sample sizes for the multinomial distribution.
Value
A named list with exactly two top-level elements:
parameters: A named list of numeric starting-value vectors for TMB:p: Fixed-effect values returned byget_fixed().re: Random-effect values returned byget_random(). This vector is empty if the model has no random effects.
These vectors follow the initialized C++ parameter order, not necessarily the row order of the input
parameterstibble. They contain values on the model's estimation scale (for example, log scale for log parameters). Parameters withestimation_type = "constant"are stored in the modules, not in these vectors. This nested list is not the input tibble.model: The instantiated C++ model module. Currently this is the catch-at-age model created withmethods::new(CatchAtAge).
The structure is:
If the returned list is assigned to initialized, pass the entire list as
fit_fims(input = initialized). For manual fitting, pass only
initialized[["parameters"]] to the parameters argument of
TMB::MakeADFun(), with random = "re" and DLL = "FIMS".
The top-level name must be parameters, not parameter_list.
No map element is returned. To map parameters when using fit_fims(), add
initialized[["map"]] as a top-level named list of factors with entries
p and/or re. Each factor must match the length and order of its
corresponding vector in initialized[["parameters"]]. Omit an entry to
leave that vector unmapped. See fit_fims() for mapping rules.
It is important that you only have one FIMS model initialized in your R
workspace at a time. Thus, after you initialize and fit the model, you should
run clear().
Examples
if (FALSE) { # \dontrun{
# Prepare data for FIMS model
data("data_big", package = "FIMS")
data_4_model <- FIMSFrame(data_big)
# Instantiate modules
parameters_list <- setup_default_parameters(data = data_4_model) |>
initialize_fims(data = data_4_model)
clear()
} # }
