Estimated N̂_{e} | Estimated m̂ | Inference efficiency | |||||||||
---|---|---|---|---|---|---|---|---|---|---|---|

L | N_{e} | H | c.v. | 95% C.I. | m | Mean | c.v. | 95% C.I. | 1 | 2 | 3 |

4 | 20 | 17.75 | 0.63 | 8.10–51.88 | 0.05 | 0.0553 | 0.51 | 0.0194–0.1070 | 0.0 | 0.2 | 99.8 |

50 | 43.32 | 0.88 | 19.21–192.77 | 0.02 | 0.0232 | 0.62 | 0.0052–0.0487 | 0.0 | 5.3 | 94.7 | |

100 | 70.02 | 1.07 | 26.91–465.71 | 0.01 | 0.0141 | 0.70 | 0.0025–0.0337 | 0.0 | 17.3 | 82.3 | |

8 | 20 | 18.75 | 0.45 | 10.18–39.06 | 0.05 | 0.0537 | 0.40 | 0.0235–0.0933 | 0.0 | 0.0 | 100.0 |

50 | 47.74 | 0.61 | 24.92–131.15 | 0.02 | 0.0212 | 0.46 | 0.0071–0.0384 | 0.0 | 0.3 | 99.7 | |

100 | 91.70 | 0.94 | 41.95–426.41 | 0.01 | 0.0110 | 0.60 | 0.0023–0.0229 | 0.0 | 5.5 | 94.5 | |

12 | 20 | 18.86 | 0.37 | 9.94–36.64 | 0.05 | 0.0530 | 0.39 | 0.0251–0.0910 | 0.0 | 0.0 | 100.0 |

50 | 48.45 | 0.48 | 26.06–107.73 | 0.02 | 0.0209 | 0.42 | 0.0085–0.0366 | 0.0 | 0.0 | 100.0 | |

100 | 98.84 | 0.80 | 47.53–332.14 | 0.01 | 0.0102 | 0.54 | 0.0028–0.0205 | 0.0 | 2.9 | 97.1 | |

16 | 20 | 18.81 | 0.36 | 10.73–34.68 | 0.05 | 0.0533 | 0.38 | 0.0260–0.0888 | 0.0 | 0.0 | 100.0 |

50 | 48.64 | 0.46 | 26.86–105.44 | 0.02 | 0.0210 | 0.41 | 0.0088–0.0366 | 0.0 | 0.0 | 100.0 | |

100 | 97.49 | 0.76 | 50.83–292.93 | 0.01 | 0.0105 | 0.49 | 0.0033–0.0205 | 0.0 | 0.8 | 99.2 |

Results are based on 1000 measures (see text) from a simulated population of size *N*_{e} receiving a proportion *m* of migrant individuals per generation. *n* is the number of diploid individuals that were sampled without replacement. Harmonic means (*H*) for *N*_{e} and arithmetic means for *m* are given for various simulated effective population sizes and various numbers of loci. Coefficients of variation (c.v.), the ratio of standard deviation over the mean, are also indicated. A 95% confidence interval (C.I.) is defined from the 5th and 95th percentiles. An assessment of inference efficiency is given with special reference to three distinct cases in the process of inference. Case 1 is the percentage of too large η_{S} values for a fit to be reliable, which is expected for a very small effective population size. Case 2 is the percentage of either negative η_{S} estimates, which give an infinite effective size, or *N̂*_{e} estimates >10 times the true *N*_{e}. Case 3 is the percentage of successful inferences from which results are given.