Skip to contents

Convert group sequential design to integer sample sizes

Source: R/gsNBCalendar.R
toInteger.Rd

Generic function to round sample sizes in a group sequential design to integers. This extends the gsDesign::toInteger() function from the gsDesign package to work with gsNB objects.

Usage

toInteger(x, ...)

# S3 method for class 'gsDesign'
toInteger(x, ratio = x$ratio, roundUpFinal = TRUE, ...)

# S3 method for class 'gsNB'
toInteger(x, ratio = x$nb_design$inputs$ratio, roundUpFinal = TRUE, ...)

Arguments

x

An object of class gsNB or gsDesign.

...

Additional arguments passed to methods.

ratio

Randomization ratio (n2/n1). If an integer is provided, rounding is done to a multiple of ratio + 1. If ratio < 1 and 1/ratio is an integer (e.g., 1:2 allocation, ratio = 0.5), rounding is done to a multiple of 1/ratio + 1. Default uses the ratio from the original design.

roundUpFinal

If TRUE (default), the final sample size is rounded up to ensure the target is met. If FALSE, rounding is to the nearest integer.

Value

An object of the same class as input with integer sample sizes.

Details

This function rounds the final sample size while maintaining the randomization ratio. When calendar analysis times are available, interim sample sizes remain expected enrollment counts at those calendar times after rescaling the accrual rate to the rounded final sample size.

When analysis_times were provided to gsNBCalendar(), expected events, exposure, and statistical information (n.I) are recomputed at each analysis time based on the new sample size and expected exposures.

Methods (by class)

  • toInteger(gsDesign): Method for gsDesign objects (calls gsDesign::toInteger()).

  • toInteger(gsNB): Method for gsNB objects.

    Rounds sample sizes in a group sequential negative binomial design to integers, respecting the randomization ratio.

Examples

nb_ss <- sample_size_nbinom(
  lambda1 = 0.5, lambda2 = 0.3, dispersion = 0.1, power = 0.9,
  accrual_rate = 10, accrual_duration = 20, trial_duration = 24
)
gs_design <- gsNBCalendar(nb_ss, k = 3, analysis_times = c(12, 18, 24))
gs_integer <- toInteger(gs_design)