Confidence intervals for misskappa coefficients

Wald confidence intervals for the coefficients of a `misskappa_estimate`, on the natural scale (`transform = "none"`) or via the variance-stabilising Fisher z transform (`transform = "fisher"`). The Fisher interval is built on the `atanh` scale with a delta-method standard error \(\mathrm{se}_z = \mathrm{se} / (1 - \hat\theta^2)\) and back-transformed with `tanh`, so it always lies in \((-1, 1)\) and tends to have better small-sample coverage near the upper boundary.

Usage

# S3 method for class 'misskappa_estimate'
confint(
  object,
  parm = NULL,
  level = 0.95,
  transform = c("none", "fisher"),
  ...
)

Arguments

object

A `misskappa_estimate` object.

parm

Optional subset of coefficients (names or indices); defaults to all.

level

Confidence level.

transform

Either `"none"` (natural-scale Wald interval) or `"fisher"` (delta-method interval on the `atanh` scale, back-transformed with `tanh`). Coefficients with \(|\hat\theta| \ge 1\) yield `NA` limits under `"fisher"`.

...

Unused; present for S3 generic conformance.

Value

A two-column numeric matrix of lower and upper limits, one row per coefficient.

Examples

fit <- kappa(dat.gwet2014, estimator = "ipw")
#> Warning: rater pair(s) rater4-rater5 co-observed by only one subject; the corresponding pairwise covariance is degenerate and the standard error unreliable.
confint(fit)                       # natural-scale Wald interval
#>                      2.5 %    97.5 %
#> Conger           0.2040066 0.6426036
#> Fleiss           0.1966366 0.6399174
#> Brennan-Prediger 0.2045079 0.6797335
confint(fit, transform = "fisher") # Fisher z interval; always within (-1, 1)
#>                      2.5 %    97.5 %
#> Conger           0.1824728 0.6162177
#> Fleiss           0.1751369 0.6133302
#> Brennan-Prediger 0.1776159 0.6470501