ci_i_one: Likelihood-Based Confidence Bound for One Parameter

In semlbci: Likelihood-Based Confidence Interval in Structural Equation Models

Likelihood-Based Confidence Bound for One Parameter

Description

Find the likelihood-based confidence bound for one parameter.

Usage

ci_i_one(
  i,
  which = NULL,
  sem_out,
  method = c("wn", "ur"),
  standardized = FALSE,
  robust = "none",
  sf_full = NA,
  sf_args = list(),
  sem_out_name = NULL,
  try_k_more_times = 0,
  ...
)

Arguments

i	The position (row number) of the target parameters as appeared in the parameter table of the lavaan::lavaan object.
which	Whether the lower bound or the upper bound is to be found. Must be "lbound" or "ubound".
sem_out	The SEM output. Currently supports lavaan::lavaan outputs only.
method	The approach to be used. Default is "wn" (Wu-Neale-2012 Method). Another method is "ur", root finding by stats::uniroot().
standardized	Logical. Whether the bound of the LBCI of the standardized solution is to be searched. Default is FALSE.
robust	Whether the LBCI based on robust likelihood ratio test is to be found. Only "satorra.2000" in lavaan::lavTestLRT() is supported for now. If "none", the default, then likelihood ratio test based on maximum likelihood estimation will be used. For "ur", "satorra.2000" is automatically used if a scaled test statistic is requested in sem_out.
sf_full	A list with the scaling and shift factors. Ignored if robust is "none". If robust is "satorra.2000" and sf_full is supplied, then its value will be used. If robust is "satorra.2000" but sf_full is NA, then scaling factors will be computed internally.
sf_args	The list of arguments to be used for computing scaling factors if robust is "satorra.2000". Used only by semlbci(). Ignored if robust is not "satorra.2000".
sem_out_name	The name of the object supplied to sem_out. NULL by default. Originally used by some internal functions. No longer used in the current version but kept for backward compatibility.
try_k_more_times	How many more times to try if the status code is not zero. Default is 0.
...	Arguments to be passed to the function corresponds to the requested method (ci_bound_wn_i() for "wn").

Details

Important Notice

This function is not supposed to be used directly by users in typical scenarios. Its interface is user-unfriendly because it should be used through semlbci(). It is exported such that interested users can examine how a confidence bound is found, or use it for experiments or simulations.

Usage

ci_i_one() is the link between semlbci() and the lowest level function (currently ci_bound_wn_i()). When called by semlbci() to find the bound of a parameter, ci_i_one() calls a function (ci_bound_wn_i() by default) one or more times to find the bound (limit) for a likelihood-based confidence interval.

Value

A list of the following elements.

• bound: The bound located. NA if the search failed.

• diags: Diagnostic information.

• method: Method used. Currently only "wn" is the only possible value.

• times: Total time used in the search.

• sf_full: The scaling and shift factors used.

• ci_bound_i_out: The original output from ci_bound_wn_i().

• attempt_lb_var: How many attempts used to reduce the lower bounds of free variances.

• attempt_more_times: How many additional attempts used to search for the bounds. Controlled by try_k_more_times.

See Also

semlbci(), ci_bound_wn_i()

Examples

data(simple_med)

library(lavaan)
mod <-
"
m ~ x
y ~ m
"
fit_med <- lavaan::sem(mod, simple_med, fixed.x = FALSE)

parameterTable(fit_med)

# Find the LBCI for the first parameter
# The method "wn" needs the constraint function.
# Use set_constraint() to generate this function:
fn_constr0 <- set_constraint(fit_med)

# Call ci_i to find the bound, the lower bound in this example.
# The constraint function, assigned to f_constr, is passed
# to ci_bound_wn_i().
# npar is an argument for ci_bound_wn_i().
out <- ci_i_one(i = 1,
                which = "lbound",
                sem_out = fit_med,
                npar = 5,
                f_constr = fn_constr0)
out$bounds

semlbci documentation built on June 22, 2024, 10:55 a.m.