In this paper we combine isotonic regression with least-squares cross-validation based model-mix method to estimate a discrete distribution with an infinite support. The new estimator is proved to be strongly consistent with $\sqrt{n}$-rate of convergence, it behaves well for moderate sized data sets and provides a trade-off between goodness-of-fit and shape constraints.

翻译：在本文中,我们结合了最小平方的跨校准模型混合法来估算离散分布,并给予无限的支持。 新的估计值被证明非常符合$\ sqrt{n}$( $) 的趋同率,它对中等尺寸的数据集表现良好,并且提供了适中和形状制约之间的权衡。