plotprofile {MQMF}

plotprofile simplifies plotting single likelihood profiles

Description

plotprofile simplifies plotting out the likelihood profiles of single parameters or variables. It is necessary to pass the function the output from the profile calculations, identifying the variable name against which to plot the likelihood while also identifying the name of the -ve log-likelihood column. Facilities are provided for defining the x and y axis labels. We need to use the function which.closest because we use a sequence of parameter values so an exact match would be highly unlikely.

Usage

plotprofile(
  prof,
  var,
  digit = c(3, 3, 3),
  xlabel = var,
  ylabel = "-ve Log-Likelihood",
  like = "-veLL",
  defpar = TRUE,
  ...
)

Arguments

prof

the results from the likelihood profile calculations. This matrix should include, as a minimum, the fixed variable of interest and the matching -ve log-likelihood in named columns.

var

the name of the variable of interest to identify the column in prof in which to find the vector of fixed values given.

digit

this is a vector of three that determine by how much the round function limits the values printed of the 95 mean at the top of the plot.

xlabel

the x-axis label, defaults to the name of the var

ylabel

the y-axis label, defaults to -ve Log-Likelihood

like

identifies the name of the column containing the -ve log-likelihood

defpar

logical, should the par values be assumed or defined, defaults to TRUE, so only one plot will be produced and any old par values will be restored afterwards. If part of a multiple plot define the formatting before calling and set this to FALSE. This will lead to the oldpar being returned invisibly so that eventually the par settings can be reset.

...

used for any other graphic parameters used.

Value

nothing but this does generate a plot.

Examples

 data(abdat)    #usually use  ~100 steps in rval, perhaps use
 rval <- seq(0.325,0.45,0.02)   # seq(0.325,0.45,0.001)
 ntrial <- length(rval)
 columns <- c("r","K","Binit","sigma","-veLL")
 result <- matrix(0,nrow=ntrial,ncol=length(columns),
                  dimnames=list(rval,columns))
 bestest <- c(r= 0.32,K=11000,Binit=4000,sigma=0.05) 
 for (i in 1:ntrial) {  #i <- 1
   param <- log(c(rval[i],bestest[2:4])) 
   parinit <- param    
   bestmodP <- nlm(f=negLLP,p=param,funk=simpspm,initpar=parinit,
                  indat=abdat,logobs=log(abdat$cpue),notfixed=c(2:4),
                  typsize=magnitude(param),iterlim=1000)
   bestest <- exp(bestmodP$estimate)
   result[i,] <- c(bestest,bestmodP$minimum)
 }
 plotprofile(result,var="r",defpar=TRUE,lwd=2)