Adaptive likelihood profiling — sbtprofile • sbt

Calculate 1D likelihood profiles with respect to single parameters or more generally with respect to linear combinations such as contrasts.

Usage

sbtprofile(
  obj,
  name,
  lincomb,
  h = 1e-04,
  ytol = 2L,
  ystep = 0.1,
  maxit = ceiling(5L * ytol/ystep),
  parm.range = c(-Inf, Inf),
  slice = FALSE,
  adaptive = TRUE,
  trace = TRUE
)

Arguments

obj: Object from MakeADFun that has been optimized.
name: Character string or integer. Name or index of a parameter to profile.
lincomb: Numeric vector. Optional linear combination of parameters to profile. By default a unit vector corresponding to name. Length must match number of parameters.
h: Numeric scalar. Initial adaptive stepsize on parameter axis. Default is 1e-04.
ytol: Numeric scalar. Tolerance that adjusts the range of the likelihood values. Default is 2.
ystep: Numeric scalar. Step size that adjusts the resolution of the likelihood profile. Default is 0.1.
maxit: Integer. Maximum number of iterations for adaptive algorithm. Default is ceiling(5 * ytol / ystep).
parm.range: Numeric vector of length 2. Valid parameter range as c(min, max). Default is c(-Inf, Inf).
slice: Logical. If TRUE, do slicing rather than profiling. Default is FALSE.
adaptive: Logical. Use adaptive step size? Default is TRUE.
trace: Logical or numeric. Trace progress? TRUE or 1 gives basic tracing, values > 1 give more information. Default is TRUE.

Value

A data frame of class tmbprofile with columns:

Parameter name: Values of the profiled parameter
value: Objective function values along the profile
Component columns: Likelihood contributions from different model components (Prior, Selectivity, Recruitment, etc.)

Details

A supplied lincomb must contain at least one nonzero element and have the same length as the active fixed-parameter vector. These requirements are assumed by the matrix construction and are not fully validated before profiling begins.