Confidence regions for qgcompmulti coefficients

Builds a bootstrap-covariance chi-squared confidence region for selected marginal structural model (MSM) coefficients from a fitted qgcomp.glm.multi() object.

Usage

confregion(object, ...)

# S3 method for class 'qgcompmulti'
confregion(
  object,
  parm = c("psi1", "psi2"),
  level = 0.95,
  method = "bootstrap_chisq",
  npoints = 200L,
  ...
)

# S3 method for class 'qgcompmulti_mi'
confregion(object, ...)

Arguments

object

A fitted object.

...

Additional arguments passed to methods.

parm

Character or integer vector identifying the MSM coefficients to include in the region. The default is c("psi1", "psi2"). Use NULL to include all MSM coefficients.

level

Confidence level for the chi-squared cutoff. Must be a single number strictly between 0 and 1.

method

Region construction method. Version 0.5.0 supports only "bootstrap_chisq", the chi-squared ellipsoid based on the bootstrap covariance matrix of the selected MSM coefficients.

npoints

Number of boundary points stored for two-parameter plotting. Ignored for regions with more than two selected coefficients.

Value

A structured qgcompmulti_region object containing the selected coefficient names, fitting-scale center, restricted bootstrap covariance matrix, confidence level, chi-squared cutoff, degrees of freedom, scale metadata, and two-dimensional plot data when applicable.

Details

Confidence regions are computed on the MSM fitting coefficient scale. For mean-difference and risk-difference estimands this is the identity scale. For odds-ratio and rate-ratio estimands this is the log coefficient scale. Version 0.5.0 does not geometrically transform confidence ellipsoids onto nonlinear ratio display scales.

The region is defined as

$$(\theta - \hat\theta)' \hat\Sigma^{-1} (\theta - \hat\theta) \leq \chi^2_{p, level},$$

where p is the number of selected coefficients and the covariance matrix is the bootstrap covariance matrix restricted to those coefficients.

Pooled multiple-imputation confidence regions are intentionally out of scope for Version 0.5.0.

Examples

if (FALSE) { # \dontrun{
dat <- sim_mixture_data(
  n = 400, pA = 3, pB = 3, rho_within_A = 0.3, rho_within_B = 0.3,
  rho_between = 0.2, psi1 = 0.5, psi2 = 0.3, psi12 = 0.2, seed = 123
)
fit <- qgcomp.glm.multi(
  f = Y ~ X1 + X2 + X3 + W1 + W2 + W3 + C,
  data = dat, mix1 = c("X1", "X2", "X3"),
  mix2 = c("W1", "W2", "W3"), B = 100, seed = 13
)
region <- confregion(fit, parm = c("psi1", "psi2"))
region
plot(region)
} # }