pwr.cohensdCI: Determine required sample size for a given confidence interval width for Cohen's d

Description

This function computes how many participants you need if you want to achieve a confidence interval of a given width. This is useful when you do a study and you are interested in how strongly twovariables are associated.

Usage

pwr.cohensdCI(d, w = 0.1, conf.level = 0.95, extensive = FALSE, silent=FALSE)

Arguments

The expected value(s) of Cohen's d.

The half-width of the desired confidence interval (the margin of error).

conf.level

The confidence level of the desired confidence interval.

extensive

Whether you want extensive results or just the sample size(s).

silent

Whether you want warnings or not.

Value

The sample size(s) you need (or, if extensive is TRUE, also the requested and resulting confidence interval bounds).

References

Peters, G. J. Y. & Crutzen, R. (forthcoming) An easy and foolproof method for establishing how effective an intervention of behavior change method is: required sample size for accurate parameter estimation in health psychology.

Maxwell, S. E., Kelley, K., & Rausch, J. R. (2008). Sample size planning for statistical power and accuracy in parameter estimation. Annual Review of Psychology, 59, 537-63. https://doi.org/10.1146/annurev.psych.59.103006.093735

Cumming, G. (2013). The New Statistics: Why and How. Psychological Science, (November). https://doi.org/10.1177/0956797613504966

Examples

Run this code


### If you want to estimate an expected effect size of
### 0.5 (tentatively qualified as a 'moderate' association),
### and you want to estimate with a maximum confidence interval
### width of .1 (e.g. from .45 to .55), you'll need a lot of people:
pwr.cohensdCI(.5, w=.05);

### If you expect a smaller effect and are content with a less
### accurate estimation, you'll need way less (but still a lot):
pwr.cohensdCI(.2, w=.1);

### If you want more extensive feedback:
pwr.cohensdCI(.2, w=.1, extensive=TRUE);

### As Cohen's d becomes larger, the confidence interval widens and
### therefore, you need more people:
pwr.cohensdCI(c(.2, .5, .8), w=.1);