This function uses bootES::bootES() to compute

pwr.bootES(data = data, ci.type = "bca", ..., w = 0.1, silent = TRUE)

Arguments

data

The dataset, as you would normally supply to bootES::bootES(); you will probably have to simulate this.

ci.type

The estimation method; by default, the default of bootES::bootES() is used ('bca'), but this is changed to 'basic' if it encounters problems.

...

Other options for bootES::bootES() (see that help page).

w

The desired 'halfwidth' of the confidence interval.

silent

Whether to provide a lot of information about progress ('FALSE') or not ('TRUE').

Value

A single numeric value (the sample size).

References

Kirby, K. N., & Gerlanc, D. (2013). BootES: An R package for bootstrap confidence intervals on effect sizes. Behavior Research Methods, 45, 905–927. doi: 10.3758/s13428-013-0330-5

Examples

### This requires the bootES package
  if (requireNamespace("bootES", quietly = TRUE)) {

  ### To estimate a mean
  x <- rnorm(500, mean=8, sd=3);
  pwr.bootES(data.frame(x=x),
             R=500,
             w=.5);

  ### To estimate a correlation (the 'effect.type' parameter is
  ### redundant here; with two columns in the data frame, computing
  ### the confidence interval for the Pearson correlation is the default
  ### ehavior of bootES)
  y <- x+rnorm(500, mean=0, sd=5);
  cor(x, y);
  requiredN <-
    pwr.bootES(data.frame(x=x,
                          y=y),
               effect.type='r',
               R=500,
               w=.2);
  print(requiredN);
  ### Compare to parametric confidence interval
  ### based on the computed required sample size
  confIntR(r = cor(x, y),
           N = requiredN);
  ### Width of obtained confidence interval
  print(round(diff(as.numeric(confIntR(r = cor(x, y),
                              N = requiredN))), 2));
}