growhamp: von Bertalanffy Growth Models for Tagging Data Incorporating...

View source: R/growhamp.R

growhampR Documentation

von Bertalanffy Growth Models for Tagging Data Incorporating Individual Variation

Description

Function fits growth models of Hampton (1991) to length and time-at-large data from tagging studies

Usage

growhamp(L1 = NULL, L2 = NULL, TAL = NULL, 
models = c(1, 2, 3, 4, 5, 6, 7), 
method = c("Nelder-Mead", "Nelder-Mead", "Nelder-Mead",
 "Nelder-Mead", "Nelder-Mead", "Nelder-Mead", "Nelder-Mead"), 
varcov = c(TRUE, TRUE, TRUE, TRUE, TRUE, TRUE, TRUE), 
Linf = list(startLinf = NULL, lowerLinf = NULL, upperLinf = NULL), 
K = list(startK = NULL, lowerK = NULL, upperK = NULL), 
sigma2_error = list(startsigma2 = NULL, lowersigma2 = NULL, uppersigma2 = NULL), 
sigma2_Linf = list(startsigma2 = NULL, lowersigma2 = NULL, uppersigma2 = NULL),
sigma2_K = list(startsigma2 = NULL, lowersigma2 = NULL, uppersigma2 = NULL), 
mu_measure = 0, sigma2_measure = 0, 
control = list(maxit = 1000))

Arguments

L1

Vector of release lengths. Each row presents the length of an individual.

L2

Vector of recapture lengths.

TAL

vector of associated time-at-large data. Calculated as the recapture date minus release date.

models

The models to fit. 1 = Faber model, 2 = Kirkwood and Somers model, 3 = Kirkwood and Somers model with model error, 4 = Kirkwood and Somers model with model and release-length-measurement error, 5 = Sainsbury model, 6 = Sainsbury model with model error, and 7 = Sainsbury model with model and release-length-measurement error. Default is all: c(1,2,3,4,5,6,7).

method

Character vector of optimization methods used in optim to solve parameters for each model. A different method can be selected for each model. Choices are "Nelder-Mead","BFGS","CG","L-BFGS-B" and "SANN". See help for optim. Default is "Nelder-Mead". If there are fewer values specified in method than the number specified in models, a warning message is produced and the last value in the method vector is used for the remaining models.

varcov

Logical vector specifying whether the parameter variance-covariance matrix of each model should be outputted. A different logical can specified for each model. If there are fewer values specified in varcov than the number specified in models, a warning message is produced and the last value in the varcov vector is used for the remaining models.

Linf

A list of starting (startLinf), lower bound (lowerLinf) and upper bound (upperLinf) of Linfinity of the von Bertalanffy equation used in the optimization. The lower and upper bounds are used only with method "L-BFGS-B".

K

A list of starting (startK), lower bound (lowerK) and upper bound (upperK) of K (growth coefficient) of the von Bertalanffy equation used in the optimization. The lower and upper bounds are used only with method "L-BFGS-B".

sigma2_error

A list of starting (startsigma2), lower bound (lowersigma2) and upper bound (uppersigma2) of the error variance used in the optimization. The lower and upper bounds are used only with method "L-BFGS-B". This parameter is used in models 1,3,4,6 and 7.

sigma2_Linf

A list of starting (startsigma2), lower bound (lowersigma2) and upper bound (uppersigma2) of the Linfinity variance used in the optimization. The lower and upper bounds are used only with method "L-BFGS-B". This parameter is used in models 2,3,4,5,6,and 7.

sigma2_K

A list of starting (startsigma2), lower bound (lowersigma2) and upper bound (uppersigma2) of the K (growth coefficient) variance used in the optimization. The lower and upper bounds are used only with method "L-BFGS-B". This parameter is used in models 5,6, and 7.

mu_measure

Release measurement error. This parameter is used in models 4 and 7. Default=0.

sigma2_measure

Variance of release measurement error. This parameter is used in models 4 and 7. Default=0.

control

A list of control parameters for optim. See function optim for details.

Details

The seven models are fitted by maximum likelihood using formulae shown in Hampton 1991. Due to the number of parameters estimated, some models can be sensitive to the initial starting values. It is recommended that the starting values are tested for sensitivity to ensure the global minimum has been reached. Sometimes, the hessian matrix, which is inverted to obtain the variance-covariance matrix, will not be positive, definite and therefore will produce an error. Again, try different starting values for parameters and lower and upper bounds if applicable.

Value

results

list element containing the parameter estimates in table format for each model. Column names are model, Linf, K, s2Linf (variance of Linf), s2K (variance of K), s2error (error variance), boundary (0 = no issues; 1 = one or more parameter estimates are at constraint boundaries), -Log Likelihood, AIC (Akaike's Information Criterion, and method

varcov

if varcov=TRUE, list element containing the variance-covariance matrix for each model.

residuals

list element containing the residuals (observed-predicted values) for each model.

Author(s)

Gary A. Nelson, Massachusetts Division of Marine Fisheries gary.nelson@mass.gov

References

Hampton, J. 1991. Estimation of southern bluefin tuna Thunnus maccoyii growth parameters from tagging data, using von Bertalanffy models incorporating individual variation. U. S. Fishery Bulletin 89: 577-590.

See Also

mort.al

Examples

## Not run: 
## Models 1,2 and 3 below are models 1,2, and 4 in Table 4.17 of ##Quinn and Deriso 
data(trout)
growhamp(L1=trout$L1,L2=trout$L2,TAL=trout$dt,models=c(1,2,3),
       method=c("Nelder-Mead","Nelder-Mead","L-BFGS-B"),
       varcov=c(TRUE,TRUE,TRUE),
       Linf=list(startLinf=650,lowerLinf=400,upperLinf=800),       
       K=list(startK=0.30,lowerK=0.01,upperK=1),
       sigma2_error=list(startsigma2=100,lowersigma2=0.1,uppersigma2=10000),
       sigma2_Linf=list(startsigma2=100,lowersigma2=0.1,uppersigma2=100000),	
       sigma2_K=list(startsigma2=0.5,lowersigma2=1e-8,uppersigma2=10))

## End(Not run)

fishmethods documentation built on April 27, 2023, 9:10 a.m.