MQMF (version 0.1.1)

fitSPM: fitSPM fits a surplus production model

Description

fitSPM fits a surplus production model (either Schaefer or Fox) by first applying optim (using Nelder-Mead) and then nlm. Being automated it is recommended that this only be used once plausible initial parameters have been identified (through rules of thumb or trial and error). It uses negLL1 to apply a negative log-likelihood, assuming log-normal residual errors and uses a penalty to prevent the first parameter (r) from becoming < 0.0. If that is not wanted then set funkone to FALSE, which would then use negLL by itself. The output object is the usual object output from nlm, which can be neatly printed using the MQMF function outfit. The $estimate values can be used in plotspmmod to plot the outcome, or in spmboot to conduct bootstrap sampling of the residuals from the CPUE model fit, to gain an appreciation of any uncertainty in the analysis. Because it requires log(parameters) it does not use the magnitude function to set the values of the parscale parameters.

Usage

fitSPM(
  pars,
  fish,
  schaefer = TRUE,
  maxiter = 1000,
  funk = simpspm,
  funkone = TRUE,
  hess = FALSE,
  steptol = 1e-06
)

Arguments

pars

the initial parameter values to start the search for the optimum. These need to be on the log-scale (log-transformed)

fish

the matrix containing the fishery data 'year', 'catch', and 'cpue' as a minimum. These exact column names are required.

schaefer

if TRUE, the default, then simpspm is used to fit the Schaefer model. If FALSE then the approximate Fox model is fitted by setting the p parameter to 1e-08 inside simpspm.

maxiter

the maximum number of iterations to be used by nlm

funk

the function used to generate the predicted cpue

funkone

default = TRUE. Means use negLL1, which constrains the first parameter (r) to be greater than 0. If FALSE then use negLL which is identical to negLL1 but lacks the constraint.

hess

default is FALSE; should one calculate the hessian matrix?

steptol

the internal step tolerance, required in case nlm reports the steptol as being too small. defaults to 1e-06

Value

an nlm output object as a list

Examples

# NOT RUN {
 data(dataspm) 
 dataspm <- as.matrix(dataspm) # faster than a data.frame
 pars <- log(c(r=0.2,K=6000,Binit=2800,sigma=0.2))
 ans <- fitSPM(pars,fish=dataspm,schaefer=TRUE,maxiter=1000)
 outfit(ans)   # Schaefer model  -12.12879
 ansF <- fitSPM(pars,dataspm,schaefer=FALSE,maxiter=1000)
 outfit(ansF)  # Fox model       -12.35283 
# }