LifeHist (version 1.0-1)

MatFit: Wrapper Function to Estimate a 2- or 3-Parameters Logistic Regression of Sexual Maturity

Description

Define the logistic model, pass the initial parameter values, the data, and the numerical optimization method(s) to estimate the model and to organize results in a list.

Usage

MatFit(p, par, matdat, method, control = ls(), itnmax)

Arguments

p

Integer, either 2 or 3, determining if the asymptotic proportion is assumed known at 1 or estimated.

par

Numeric vector of 2 or 3 components with initial paramater values.

matdat

A data.frame where each row is an individual fish, and the columnds are a continuous predictor, and ordinal keys for sex, month, and maturity stage.

method

Character or character vector, defining the list of numerical methods to use. See help pages for function optimx().

control

List, optimization control parameters to be passed to optimx. See help for function optimx().

itnmax

Integer, maximum number of iterations, to be passed to optimx().

Value

model

Type of model, matlik.2p or matlik.3p

method

Name of numerical optimization method

converg

Convergence message

kkt

The Karush-Kuhn-Tucker optimality conditions

AIC

The Akaike Information Criterion

pars

Maximum likelihood estimates of model parameters

num.grads

Numerical gradients at the maximum likelihood estimates

stdev

Estimated standard deviations of maximum likelihood estimates of model parameters

Cor

The estimated correlation matrix of maximum likelihood estimates of model parameters

Details

The function controls the estimation process. From a valid value of p, it will define the proper process model, either 2-parameters or 3-parameters logistic regression. The 3-parameter model might be useful outside the reproductive season when not all adult individuals are ready to reproduce, whereas the 2-parameter model assumes that above certain value of the continuous predictor all individuals are reproducing.

The function will re-organize optimx()'s output by adding some items (AIC, standard errors, correlation matrix) and ignoring other items. The output is a list of results in a list of optimization methods.

Examples

data(BlackAngler.lenmatdat)
BA.matlen     <- MatCount(matdat=BlackAngler.lenmatdat,
                          fem.key=2,
                          mal.key=1,
                          stage.key=2,
                          season.key=c(5,7))
plot(BA.matlen[[2]],
     pred.50=35,
     pred.95=55,
     pred.unit="Length (cm)",
     top.text="Males Rep Season",
     lwd=2, xlab="", ylab="")
par(mfrow=c(2,2), mar=c(4,4,1,1), oma=c(4,4,1,1))
plot(BA.matlen[[1]],pred.50=35,pred.95=55,pred.unit="Length (cm)",
     top.text="Males NoRep Season", lwd=2, xlab="", ylab="")
plot(BA.matlen[[2]],pred.50=35,pred.95=55,pred.unit="Length (cm)",
     top.text="Males Rep Season", lwd=2, xlab="", ylab="")
plot(BA.matlen[[3]],pred.50=85,pred.95=110,pred.unit="Length (cm)",
     top.text="Females NoRep Season", lwd=2, xlab="", ylab="")
plot(BA.matlen[[4]],pred.50=55,pred.95=75,pred.unit="Length (cm)",
     top.text="Females Rep Season", lwd=2, xlab="", ylab="")
require(optimx)
BA.matlen.mal <- MatFit(p=2,
                        par=c(35,55),
                        matdat=BA.matlen[[2]],
                        method=c("spg", "CG", "Nelder-Mead"),
                        itnmax=100)
#
plot(BA.matlen[[2]],
     pred.50=BA.matlen.mal[[1]]$par.mle[1],
     pred.95=BA.matlen.mal[[1]]$par.mle[2],
     pred.unit="Length (cm)",
     top.text="Males Reproductive Season",
     lwd=2,
     xlab="Length (cm)",
     ylab="Proportion Mature")