Skip to contents

ODA minimum sample size via bisection

Source: R/oda_power.R
oda_sample_size.Rd

Finds the minimum per-group sample size \(n\) (balanced design) at which power reaches or exceeds power_target. Uses bisection over oda_power() with a fixed RNG seed for stable search.

Usage

oda_sample_size(
  power_target = 0.8,
  p1 = NULL,
  p2 = NULL,
  ess = NULL,
  alpha = 0.05,
  comp = 1L,
  nsim = 10000L,
  mc_seed = 42L,
  n_min = 2L,
  n_max = 2000L
)

Arguments

power_target

Target power. Default 0.80.

p1

Probability of the event in class 0. Ignored when ess is supplied.

p2

Probability of the event in class 1. Ignored when ess is supplied.

ess

Effect Strength for Sensitivity (percent, \(0 < \text{ESS} < 100\)). Mutually exclusive with p1/p2.

alpha

Nominal significance level. Default 0.05.

comp

Number of comparisons for Sidak correction. Default 1.

nsim

Number of Monte Carlo replications per candidate \(n\). Default 10000.

mc_seed

Integer seed used for every oda_power() call during the bisection. Using a fixed seed yields stable, reproducible results. Default 42L.

n_min

Minimum \(n\) to search. Default 2.

n_max

Maximum \(n\) to search. If power at n_max is still below power_target, an error is raised. Default 2000.

Value

An object of class "oda_sample_size", a list with elements:

n

Minimum per-group sample size achieving power_target.

power_achieved

Estimated power at n.

power_target

Input target power.

p1, p2, ess_input

Effect-size inputs.

alpha, alpha_adj, comp

Alpha parameters.

nsim, mc_seed

Simulation parameters.

Details

Scope: unit-weighted, binary class, binary (2-level) attribute only. This is not a general CTA, LORT, SDA, weighted, or multiclass sample-size method. For unbalanced designs, call oda_power() directly across a candidate grid.

References

Rhodes, N. J. (2020). Statistical power analysis in ODA, CTA and Novometrics. Optimal Data Analysis, 9. https://odajournal.files.wordpress.com/2020/02/v9a5.pdf

Yarnold PR, Soltysik RC (2005). Optimal Data Analysis: A Guidebook with Software for Windows. Washington, DC: APA Books.

Examples

# Minimum n for ESS = 48%, 80% power (use nsim >= 500L for publication-quality estimates)
oda_sample_size(ess = 48, nsim = 200L, mc_seed = 42L)
#> ODA Sample Size Analysis
#>   Effect:  p1 = 0.2600, p2 = 0.7400  (ESS = 48.00%)
#>   alpha = 0.0500
#>   Target power:    0.800
#>   Minimum n/group: 22
#>   Power achieved:  0.850
#>   nsim = 200

# \donttest{
# 90% power target (publication-quality nsim)
oda_sample_size(ess = 48, power_target = 0.90, nsim = 500L, mc_seed = 42L)
#> ODA Sample Size Analysis
#>   Effect:  p1 = 0.2600, p2 = 0.7400  (ESS = 48.00%)
#>   alpha = 0.0500
#>   Target power:    0.900
#>   Minimum n/group: 23
#>   Power achieved:  0.902
#>   nsim = 500
# }