Back to Index

Conversion between some effect sizes

Conversion between some effect sizes
Conversion between some effect sizes

Perugini, M., et al. (2018). A Practical Primer To Power Analysis for Simple Experimental Designs. International Review of Social Psychology, 31(1): 20, 1–23, DOI: https://doi.org/10.5334/irsp.181

Design

3x2 Mixed ANOVA (3 between x 2 within)

Factors: SES origin (poor-middle-rich) x Time (pre-post vignette). Six groups.

G*Power

G*Power settings:

  • Test family: F tests
  • Statistical test: ANOVA: Repeated measures, within-between interaction
  • Number of groups: 3
  • Number of measurements: 2
  • Corr among rep measures: 0.5
  • Nonsphericity corection: 1
  • Power: 0.9
Cohen’s f Total Sample Size
0.25 54
0.20 84
0.15 144
0.10 321
  • Power: 0.85
Cohen’s f Total Sample Size
0.25 48
0.20 72
0.15 126
0.10 279
  • Power: 0.80
Cohen’s f Total Sample Size
0.25 42
0.20 66
0.15 111
0.10 246

Mediation

Using https://schoemanna.shinyapps.io/mc_power_med/ Schoemann’s online tool for a Monte Carlo power analysis with 20,000 samples for a simple mediation model. Since our model involves multicategorical mediation with three groups in the antecedent variable, we multiply the sample size by 1.5 .

Input parameters:

  • The correlation between conditions and mediator (a path), as well as between mediator and outcome (b path), is expected to be medium-sized (r=.24)
  • The correlation between conditions and outcome (c’ path) is expected to be small (r=.1).
Power Sample size (2 groups) Sample size (3 groups)
0.80 195 293
0.85 215 323
0.90 240 360
0.95 280 420
  • The correlation between conditions and mediator (a path), as well as between mediator and outcome (b path), is expected to be medium-sized (r=.3)
  • The correlation between conditions and outcome (c’ path) is expected to be small (r=.1).
Power Sample size (2 groups) Sample size (3 groups)
0.80 125 188
0.85 135 203
0.90 145 218
0.95 165 248

Preregistration statement

Based on prior research and power analyses, we aim to collect usable data from 287 participants. To account for possible exclusions due to incorrect responses to an attention-check question, we will add approximately 5% to this number, resulting in a target of 300 participants.