confint.admbsecr: Confidence intervals for admbsecr model parameters

In b-steve/admbsecr: SECR models with supplementary location information

Description

Computes confidence intervals for one or more parameters estimated in an admbsecr model object.

Usage

## S3 method for class 'admbsecr'
confint(object, parm = "fitted", level = 0.95,
  linked = FALSE, ...)

## S3 method for class 'admbsecr.boot'
confint(object, parm = "fitted", level = 0.95,
  method = "default", linked = FALSE, qqplot = FALSE, ask = TRUE, ...)

Arguments

object	A fitted model from admbsecr.
parm	A character vector containing either parameter names, or a subset of "all", "derived", "fitted", and "linked"; "fitted" corresponds to the parameters of interest, "derived" corresponds to quantities that are functions of these parameters (e.g., the effective survey area or animal density from an acoustic survey), and "linked" corresponds to the parameters AD Model Builder has maximised the likelihood over.
level	the confidence level required.
linked	Logical, if TRUE, intervals for fitted parameters are calculated on their link scales, then transformed back onto their "real" scales.
...	Other parameters (for S3 generic compatibility).
method	A character string specifying the method used to calculate the confidence intervals. See 'Details' below.
qqplot	Logical, if TRUE and method is "default" then a normal QQ plot is plotted. The default method is based on a normal approximation; this plot tests its validity.
ask	Logical, if TRUE, hitting return will show the next plot.

Details

Options for the argument method are as follows: "default" for intervals based on a normal approximation using the calculated standard errors (for objects of class admbsecr.boot, these standard errors are calculated from the bootstrap procedure); "default.bc" is a bias-corrected version of default, whereby the estimated bias is subtracted from each confidence limit; "basic" for the so-called "basic" bootstrap method; and "percentile" for intervals calculated using the bootstrap percentile method (the latter three are only available for objects of class admbsecr.boot; see Davison & Hinkley, 1997, for details).

For method "default" with objects of class admbsecr.boot, the appropriateness of the normal approximation can be evaluated by setting qqnorm to TRUE. If this indicates a poor fit, set linked to TRUE and evaluate the QQ plot to see if this yields an improvement (see Davison & Hinkley, 1997, pp. 194, for details).

References

Davison, A. C., and Hinkley, D. V. (1997) Bootstrap methods and their application. Cambridge: Cambridge University Press.

b-steve/admbsecr documentation built on May 11, 2019, 5:20 p.m.