Profile an assessment parameter

Fixes a scalar parameter or a mapped parameter group at each supplied value, then re-optimises the remaining parameters using the usual two-pass fit.

Usage

opal_profile(
  x,
  parameter,
  values,
  element = 1L,
  fit_args = list(),
  name = "profile"
)

Arguments

x

A fitted opal_obj.

parameter

Name of a parameter block, on its stored scale.

values

Distinct finite profile values on that scale.

element

One-based element within the block, in R column order.

fit_args

Named arguments passed to opal_fit().

name

Name of the stored result.

Value

An updated object containing a table, component contributions, conditional fits, and failed-point messages, accessible with opal_derived().

Details

This profiles the complete penalised objective, including priors and process penalties; it is a likelihood profile only when those terms are absent. Fixed effects can be profiled in a Laplace model, but random effect blocks cannot. Shared map elements are fixed together. The original fit is the reference; a negative objective difference indicates that a conditional fit improved on it. Failed points are never interpolated.

See also

Other assessment sensitivity: opal_grid(), opal_grid_draws(), opal_grid_mcmc(), plot_opal_profile()