排列上的指数族估计
概率论
2016-05-05 v3
摘要
针对一类包含带有 Spearman 足距（Spearman's Footrule）和 Spearman 秩相关统计量的 Mallows 模型的单参数排列指数族，计算了其归一化常数的渐近性。证明了最大似然估计（MLE）及其一种可计算的近似具有一致性。Besag 的伪似然估计被证明是 -一致的。证明了一种迭代算法（IPFP）收敛于极限归一化常数。此外，还分析了带有 Kendall's Tau 的 Mallows 模型，以展示本文工具的灵活性。
引用
@article{arxiv.1307.0978,
title = {Estimation in exponential families on permutations},
author = {Sumit Mukherjee},
journal= {arXiv preprint arXiv:1307.0978},
year = {2016}
}
备注
The presentation of the paper is changed. Proof of consistency of MLE and an approximation to the MLE is included. The convergence of the discrete IPFP algorithm is analyzed as opposed to the continuous version