ParameterizeRecaModels: Parameterize Reca.

In StoXProject/RstoxFDA: Fisheries Dependent Analysis with RstoX

Parameterize Reca.

Description

Runs estimation of parameters for Reca models. Invokes eca.estimate.

Usage

ParameterizeRecaModels(
  RecaData,
  Nsamples = integer(),
  Burnin = integer(),
  Thin = integer(),
  ResultDirectory = character(),
  Lgamodel = c("log-linear", "non-linear"),
  Delta.age = numeric(),
  Seed = numeric(),
  UseCachedData = FALSE
)

Arguments

RecaData	RecaData as returned from PrepareRecaEstimate
Nsamples	number of MCMC samples that will be made available for eca.predict. See documentation for eca.estimate,
Burnin	number of MCMC samples run and discarded by eca.estimate before any samples are saved. See documentation for eca.estimate.
Thin	controls how many iterations are run between each samples saved. Defaults to 0. This may be set to account for autocorrelation introduced by Metropolis-Hastings simulation. see documentation for eca.estimate
ResultDirectory	a directory where Reca may store temp-files eca.estimate and eca.predict. See details.
Lgamodel	The length age relationship to use for length-age fits (options: "log-linear", "non-linear": Schnute-Richards model). See documentation for eca.estimate. Defaults to log-linear
Delta.age	see documentation for eca.estimate. Defaults to 0.001.
Seed	see documentation for eca.estimate. Defaults to random seed.
UseCachedData	if TRUE Parameterization is not run, but any previous runs for exactly the same arguments are returned.

Details

eca.estimate performs Markov-chain Monte Carlo (MCMC) simulations to determine maximum likelihood of parameters for the given samples. This is computationally intensive and run time may be noticable. For a given model configuration running time is mainly determined by the parameters 'Nsample', 'Burnin' and 'Thin'.

Seed

If 'Seed' is not provided a random seed is chosen. This is stored in the returned data (RecaParameterData$GlobalVariables$Seed). This seed is passed to Reca, but not all versions of Reca has provided exact reproducability for a given seed, so the behaviour is dependent on the installed Reca-version.

Caching

The argument 'UseCachedData' allows previously computed parameterization to be returned in stead of parameterizing again. If no previous run is located in the 'ResultDirectory', or the arguments or data that are passed to Reca differs from the previous run, execution will halt with an error when 'UseCachedData'. In this respect it may also be useful to note a counter intuitive aspect of the argument 'Seed'. If 'Seed' was not provided for the previous run, the arguments will be considered equal if the seed is set to the value returned on the previous run (RecaParameterData$GlobalVariables$Seed).

ResultDirectory files:

Various report functions may use output of this function with the function eca.predict which samples the posterior distributions of parameters. Communication between eca.estimate and eca.predict is managed by writing and reading files, and a directory for storing intermediate calculations must be provided with the parameter 'ResultDirectory'. For multi-chain analysis, a different directory should be provided for each chain. The result directory will be created if it does not exist.

Be aware that this breaks with the general design of StoX and somewhat limits the transferrability of StoX projects between computers.

Value

RecaParameterData results from Reca Model Parameterization.

See Also

PrepareRecaEstimate for model configuration, and data preparation for this function, and RunRecaModels for obtaining predictions / estimates from the Reca-models.

StoXProject/RstoxFDA documentation built on Jan. 13, 2024, 12:32 a.m.