Skip to contents

Calculate confidence intervals

Source: R/methods.R
confint.glmmTMB.Rd

Calculate confidence intervals

Usage

# S3 method for class 'glmmTMB'
confint(
  object,
  parm = NULL,
  level = 0.95,
  method = c("wald", "Wald", "profile", "uniroot"),
  component = c("all", "cond", "zi", "other"),
  estimate = TRUE,
  include_nonest = FALSE,
  parallel = c("no", "multicore", "snow"),
  ncpus = getOption("profile.ncpus", 1L),
  cl = NULL,
  full = FALSE,
  ...
)

Arguments

object

glmmTMB fitted object.

parm

which parameters to profile, specified

  • by index (position) [after component selection for confint, if any]

  • by name (matching the row/column names of vcov(object,full=TRUE))

  • as "theta_" (random-effects variance-covariance parameters), "beta_" (conditional and zero-inflation parameters), or "disp_" or "sigma" (dispersion parameters)

Parameter indexing by number may give unusual results when some parameters have been fixed using the map argument: please report surprises to the package maintainers.

level

Confidence level.

method

'wald', 'profile', or 'uniroot': see Details function)

component

Which of the three components 'cond', 'zi' or 'other' to select. Default is to select 'all'.

estimate

(logical) add a third column with estimate ?

include_nonest

include dummy rows for non-estimated (mapped, rank-deficient) parameters?

parallel

method (if any) for parallel computation

ncpus

number of CPUs/cores to use for parallel computation

cl

cluster to use for parallel computation

full

CIs for all parameters (including dispersion) ?

...

arguments may be passed to profile.glmmTMB (and possibly from there to tmbprofile) or tmbroot

Details

Available methods are

"wald"

These intervals are based on the standard errors calculated for parameters on the scale of their internal parameterization depending on the family. Derived quantities such as standard deviation parameters and dispersion parameters are back-transformed. It follows that confidence intervals for these derived quantities are typically asymmetric.

"profile"

This method computes a likelihood profile for the specified parameter(s) using profile.glmmTMB; fits a spline function to each half of the profile; and inverts the function to find the specified confidence interval.

"uniroot"

This method uses the uniroot function to find critical values of one-dimensional profile functions for each specified parameter.

At present, "wald" returns confidence intervals for variance parameters on the standard deviation/correlation scale, while "profile" and "uniroot" report them on the underlying ("theta") scale: for each random effect, the first set of parameter values are standard deviations on the log scale, while remaining parameters represent correlations on the scaled Cholesky scale. For a random effects model with two elements (such as a random-slopes model, or a random effect of factor with two levels), there is a single correlation parameter \(\theta\); the correlation is equal to \(\rho = \theta/\sqrt{1+\theta^2}\). For random-effects terms with more than two elements, the mapping is more complicated: see https://github.com/glmmTMB/glmmTMB/blob/master/misc/glmmTMB_corcalcs.ipynb

Examples

data(sleepstudy, package="lme4")
model <- glmmTMB(Reaction ~ Days + (1|Subject), sleepstudy)
model2 <- glmmTMB(Reaction ~ Days + (1|Subject), sleepstudy,
    dispformula= ~I(Days>8))
confint(model)  ## Wald/delta-method CIs
#>                                  2.5 %    97.5 %  Estimate
#> (Intercept)                 232.773317 270.03687 251.40509
#> Days                          8.895915  12.03866  10.46729
#> Std.Dev.(Intercept)|Subject  25.357109  51.14422  36.01207
confint(model,parm="theta_")  ## Wald/delta-method CIs
#>                                2.5 %   97.5 % Estimate
#> Std.Dev.(Intercept)|Subject 25.35711 51.14422 36.01207
confint(model,parm=1,method="profile")
#>                2.5 %   97.5 %
#> (Intercept) 231.9922 270.8178