Data set | Recombine | infs | |
---|---|---|---|

1 | 13 | 26 | 57 |

2 | 9 | 21 | 24 |

3 | 24 | 27 | 30 |

4 | 10 | 29 | 26 |

5 | 10 | 27 | 34 |

6 | 5 | 33 | 42 |

7 | 4 | 25 | 45 |

8 | 8 | 23 | 38 |

9 | 7 | 53 | 50 |

10 | 12 | 27 | 29 |

Median | 10 | 27 | 36 |

0.62 | 0.17 | 0.11 |

Both infs and Recombine were run with θ fixed at its true value. infs was run for 20,000 iterations with five driving values for (10, 30, 40, 50, and 60). The effective sample size (ESS) at the MLE is always <4, indicating that infs had very little confidence in its estimated likelihood curve (and the estimated 95% C.I.'s failed to include the true in all but one case). Recombine was run with five short runs of 20,000 iterations and one long run of 1 million iterations, using three heating temperatures, initializing the runs at the true value of , with θ fixed as 10. The CPU times (on an 800-MHz Pentium III processor) for data set 8 were ~30 hr for infs and Recombine and 30 sec for .