Check Convergence — Converge • MSEtool Functions

Have I undertaken enough simulations (nsim)? Has my MSE converged on stable (reliable) performance metrics?

Usage

Converge(
  MSEobj,
  PMs = c("Yield", "P10", "AAVY"),
  maxMP = 15,
  thresh = 0.5,
  ref.it = 20,
  inc.leg = FALSE,
  all.its = FALSE,
  nrow = NULL,
  ncol = NULL,
  silent = FALSE
)

Arguments

MSEobj: An MSE object of class 'MSE'
PMs: A character vector of names of the PM methods or a list of the PM methods
maxMP: Maximum number of MPs to include in a single plot
thresh: The convergence threshold. Maximum root mean square deviation over the last ref.it iterations
ref.it: The number of iterations to calculate the convergence statistics. For example, a value of 20 means convergence diagnostics are calculated over last 20 simulations
inc.leg: Logical. Should the legend be displayed?
all.its: Logical. Plot all iterations? Otherwise only (nsim-ref.it):nsim
nrow: Numeric. Optional. Number of rows
ncol: Numeric. Optional. Number of columns
silent: Hide the messages printed in console?

Value

A table of convergence results for each MP

Details

Performance metrics are plotted against the number of simulations. Convergence diagnostics are calculated over the last ref.it (default = 20) iterations. The convergence diagnostics are:

Is the order of the MPs stable over the last ref.it iterations?
Is the average difference in performance statistic over the last ref.it iterations < thresh?

By default three commonly used performance metrics are used:

Average Yield Relative to Reference Yield
Probability Spawning Biomass is above 0.1BMSY
Probability Average Annual Variability in Yield is < 20 per cent

Additional or alternative performance metrics objects can be supplied. Advanced users can develop their own performance metrics.

Author

A. Hordyk

Examples

if (FALSE) {
MSE <- runMSE()
Converge(MSE)
}