Skip to contents

Confidence intervals for fitted INAD models

Source: R/ci_inad.R
ci_inad.Rd

Computes confidence intervals for selected parameters from a fitted INAD model. For the fixed effect case, Wald intervals for time varying alpha and theta are computed via Louis identity for supported thinning-innovation combinations. For block effects tau, profile likelihood intervals are computed by fixing one component of tau and re maximizing the log likelihood over nuisance parameters. For negative binomial innovations, Wald intervals for the innovation size parameter are computed using a one dimensional observed information approximation per time point, holding other parameters fixed at their fitted values.

Usage

ci_inad(
  y,
  fit,
  blocks = NULL,
  level = 0.95,
  idx_time = NULL,
  ridge = 0,
  profile_maxeval = 2500,
  profile_xtol_rel = 1e-06
)

Arguments

y

Integer matrix with n_subjects rows and n_time columns.

fit

A fitted model object returned by fit_inad.

blocks

Optional integer vector of length n_subjects. Required for block effect intervals. If provided, should match fit$settings$blocks.

level

Confidence level between 0 and 1.

idx_time

Optional integer vector of time indices for which to compute intervals. Default is all time points.

ridge

Nonnegative ridge value added to the observed information matrix used for Louis based Wald intervals.

profile_maxeval

Maximum number of function evaluations used in the profile likelihood refits.

profile_xtol_rel

Relative tolerance used in the profile likelihood refits.

Value

An object of class inad_ci, a list with elements settings, level, alpha, theta, nb_inno_size, and tau. Each non NULL interval element is a data frame with columns param, est, lower, upper, and possibly se and width.

Examples

if (FALSE) { # \dontrun{
fit <- fit_inad(y, order = 1, thinning = "nbinom", innovation = "bell", blocks = blocks)
ci <- ci_inad(y, fit, blocks = blocks)
ci$alpha
ci$theta
ci$tau
} # }