Figure | Panel | Data structure | Type of test | Power |
---|---|---|---|---|

2 | C | Normality not assumed | Mann-Whitney | 55.9 |

D | Normality not assumed | Mann-Whitney | 53.2 | |

F | Normality not assumed | Mann-Whitney | 61.7 | |

G | Normality not assumed | Mann-Whitney | 10.7 | |

H | Normality not assumed | Mann-Whitney | 12.3 | |

I | Normality not assumed | Mann-Whitney | 2.6 | |

J | Normality not assumed | Mann-Whitney | 12.6 | |

3 | B | Normality not assumed | Mann-Whitney | 21.5 |

C | Normality not assumed | Mann-Whitney | 63.3 | |

D | Normality not assumed | Mann-Whitney | 36.4 | |

4 | B | Normality not assumed | Mann-Whitney | 100 |

D | Normality not assumed | Mann-Whitney | 100100 | |

7 | A | Normality not assumed | Mann-Whitney | 37.1 |

B | Normality not assumed | Mann-Whitney | 90 | |

J | Normality not assumed | Mann-Whitney | 100100 | |

K | Normality not assumed | Mann-Whitney | 99.9100 |

From Fisher's χ

^{2}test for combined probabilities, we have that χ^{2}(df = 2*m*) ∼ –2ln(*p*1p2…*pm*), where*pi*is the*p*value for the*i*th independent experiment. The*post hoc*expected value of χ^{2}is just the df + the noncentrality parameter (λ). Thus, for three independent experiments, say λ is given by –2ln(p1p2p3) – 6. The power can then be obtained directly using G*Power 3 software with α = 0.05/k.