RMark (version 3.0.0)

get.real: Extract or compute sets of real parameters

Description

Extracts or computes real parameters for a particular type of parameter (parameter) and returns in either a table (dataframe) format or in PIM format.

Usage

get.real(
  model,
  parameter,
  beta = NULL,
  se = FALSE,
  design = NULL,
  data = NULL,
  vcv = FALSE,
  show.fixed = TRUE,
  expand = FALSE,
  pim = TRUE
)

Value

estimates: if se=FALSE and Beta=NULL, a matrix of estimates or list of matrices for more than one group, and if se=TRUE or beta=is not NULL and vcv=FALSE a dataframe of estimates with attached design data. If vcv=TRUE, a list is returned with elements vcv.real and the dataframe estimates as returned with se=TRUE.

Arguments

model

MARK model object

parameter

type of parameter in model (character) (e.g.,"Phi")

beta

values of beta parameters for computation of real parameters

se

if TRUE uses table format and extracts se and confidence intervals, if FALSE uses PIM format with estimates only

design

a numeric design matrix with any covariate values filled in with numerical values

data

covariate data to be averaged for estimates if design=NULL

vcv

if TRUE computes and returns the v-c matrix of the subset of the real parameters

show.fixed

if TRUE fixed values are returned rather than NA in place of fixed values

expand

if TRUE, returns vcv matrix for unique parameters and only simplified unique parameters if FALSE

pim

if TRUE and se=FALSE, returns a list of estimates in PIM format

Author

Jeff Laake

Details

This function is called by summary.mark to extract(from model$results$real) sets of parameters for display. But, it can also be useful to compute particular sets of real parameters for output, manipulation or plotting etc. It is closely related to compute.real and it uses that function when it computes (rather than extracts) real parameters. It provides an easier way to extract/compute real estimates of a particular type (parameter).

The real parameter estimates are computed when either 1) model$chat > 1, 2) design, data, or beta are specified with non-NULL values for those arguments, or 3) vcv=TRUE. If none of the above hold, then the estimates are extracted.

If se=FALSE and estimates are shown in triangular or square PIM structure depending on the model and parameter. For triangular, the lower half of the matrix is shown as NA (not applicable in this case). If se=TRUE, the estimate, standard error and confidence interval for the real parameters with the accompanying design data are combined into a dataframe.

If the model contains individual covariates, there are 3 options for specifying the covariate values that are used for the real estimates. If neither design nor data are specified, then the real estimates are computed with the average covariate values used in the MARK output. This is what is done in the call from summary.mark. Alternatively, the argument design can be given a numeric design matrix for the model with the covariates given specific values. This can be done with find.covariates and fill.covariates. Finally, a quicker approach is to specify a dataframe for data which is averaged for the numeric covariate values and these are automatically filled into the design matrix of the model with calls to find.covariates and fill.covariates from compute.real which is used for computation. The second and third options are essentially the same but creating the completed design matrix will be quicker if multiple calls are made with the same completed design matrix for different parameters. The dataframe for data can contain a single entry to specify certain values for computation.

See Also

summary.mark,compute.real

Examples

data(example.data)
pregion=list(formula=~region)
PhiAge=list(formula=~Age)
mod=mark(example.data,model.parameters=list(p=pregion,Phi=PhiAge),
 groups=c("sex","age","region"),age.var=2,initial.ages=c(0,1,2),threads=1,delete=TRUE)
# extract list of Phi parameter estimates for all groups in PIM format 
Phi.estimates=get.real(mod,"Phi")  
# print out parameter estimates in triangular PIM format
for(i in 1:length(Phi.estimates))  
{
  cat(names(Phi.estimates)[i],"\n")
  print(Phi.estimates[[i]]$pim,na.print="")
}
require(plotrix)
#extract parameter estimates of capture probability p with se and conf intervals
p.table=get.real(mod,"p",se=TRUE) 
print(p.table[p.table$region==1,])  # print values from region 1
estimates=by(p.table$estimate,p.table$region,mean)
lcl=by(p.table$lcl,p.table$region,mean)
ucl=by(p.table$ucl,p.table$region,mean)
plotCI(c(1:4),estimates,ucl-estimates,estimates-lcl,xlab="Region",
         ylab="Capture probability",
		ylim=c(.5,1),main="Capture probability estimates by region")