Skip to contents

Add simultaneous confidence intervals to nomatchfit object

Source: R/add_simultaneous_ci.R
add_simultaneous_ci.Rd

Computes simultaneous confidence intervals, which maintain the specified coverage level for the parameter of interest across all evaluation timepoints jointly. This is useful for making inferences about the entire cumulative incidence or effectiveness curve. The simultaneous confidence intervals returned contrast with the pointwise confidence intervals automatically returned with a nomatchfit object when boot_reps > 0.

Usage

add_simultaneous_ci(object, seed = NULL)

Arguments

object

An object of class nomatchfit created by nomatch() or matching(). Must

  • contain evaluations at at least 2 timepoints (length(object$timepoints) > = 2),

  • contain more than one bootstrap replicate (boot_reps > 1)

  • contain non-null bootstrap samples (keep_boot_samples = TRUE when fitting).

seed

Integer seed to ensure reproducible critical values for simultaneous confidence intervals (relied on random number generation). Default is NULL (no seed set).

Value

The original nomatchfit object with the following modifications:

estimates

The matrix for each term in the estimates list gets additional columns describing the simultaneous confidence interval bounds and construction: simul_lower, simul_upper, simul_n

simul_z_star

Named vector containing the critical values used for each term

simul_excluded_timepoints

Named list containing a vector of timepoints excluded from the simultaneous bands for each term due to insufficient bootstrap samples

Details

Critical values are computed using the bootstrap covariance structure across timepoints. If any timepoint has more than 5% missing bootstrap samples, it is excluded from the simultaneous band and a warning is issued.

Examples

# Fit model with bootstrap samples
fit <- nomatch(
  data = simdata,
  outcome_time = "Y",
  outcome_status = "event",
  exposure = "V",
  exposure_time = "D_obs",
  covariates = c("x1", "x2"),
  timepoints = seq(30, 180, by = 30),
  immune_lag = 14,
  boot_reps = 100,         # must be > 1
  keep_boot_samples = TRUE # must be set to TRUE (default)
)
#> Bootstrapping 100 samples...
#> Bootstrap completed in 32.12 secs

# Add simultaneous CIs
fit_simul <- add_simultaneous_ci(fit, seed = 123)

# Look at results
fit_simul
#> 
#>  Risk Ratio Estimates 
#> ================================================== 
#> Call: nomatch(data = simdata, outcome_time = "Y", outcome_status = "event", 
#>     exposure = "V", exposure_time = "D_obs", covariates = c("x1", 
#>         "x2"), immune_lag = 14, timepoints = seq(30, 180, by = 30), 
#>     boot_reps = 100, keep_boot_samples = TRUE) 
#> 
#> Result:
#>   Timepoint Estimate 95% Wald CI: Lower 95% Wald CI: Upper Wald p-value
#> 1        30    0.534              0.346              0.826     4.81e-03
#> 2        60    0.605              0.490              0.746     2.74e-06
#> 3        90    0.603              0.511              0.712     2.28e-09
#> 4       120    0.661              0.561              0.779     7.28e-07
#> 5       150    0.731              0.629              0.849     4.28e-05
#> 6       180    0.828              0.715              0.959     1.18e-02
#>   95% Simul CI: Lower 95% Simul CI: Upper
#> 1               0.308               0.928
#> 2               0.463               0.789
#> 3               0.489               0.744
#> 4               0.537               0.813
#> 5               0.604               0.884
#> 6               0.687               0.997
#> 
#> Use summary() for more details
#> Use plot() to visualize results

# Visualize
plot(fit_simul, ci_type = "simul")