Maximum entropy sampling
Maximum entropy sampling
复制标题
DOI：
10.1080/02664768700000020
复制
发表时间：
1987
影响因子：
1.5
通讯作者：
M. Shewry;H. Wynn
中科院分区：
文献类型：
--
作者：
M. Shewry;H. Wynn
在最佳或高效数据收集的历史上有几个不同的研究方向。在实验设计领域，英国学派从 R. A. Fisher 和 F. Yates 的农业田间试验工作开始，并产生了大量组合设计方面的文献。美国决策理论学派是在 A. Wald 奠定的统计设计理论基础上发展起来的。其中，出现了两条截然不同的思路：J. Wolfowitz 和 J. Kiefer 的优化设计理论以及 D. Blackwell、L. Le Cam 和 DV Lindley 的信息论方法。这些最近在贝叶斯实验设计的旗帜下进行了研究（详细讨论参见 Chaloner，1984）。更广泛的数据收集领域必须包括调查抽样。空间抽样可以被认为是调查抽样和优化设计思想的融合。最近的一篇文章 Ylvisaker (1986) 汇集了许多想法。在本文中，提出了一个简单的标准和示例，试图理解受控实验和更被动的观察之间的范围。 Wynn (1982a) 中已经通过一些哲学参考文献讨论了这种对比。我们希望阐明的另一个区别是对有限、封闭系统的观察与对无限或更广泛环境的观察之间的区别。
There are several distinct strands of research in the history of optimum or efficient data collection. Within the area of experimental design the British School started with the work of R. A. Fisher & F. Yates in agricultural field trials and led to a huge literature in combinatorial design. The American decision theoretic school developed out of the foundations of statistical design theory laid down by A. Wald. Within this, two rather distinct strands emerged: the theory of optimum design of J. Wolfowitz & J. Kiefer and the information theory approach of D. Blackwell, L. Le Cam and DV Lindley. These have been studied recently under the banner of Bayesian design of experiments (see Chaloner, 1984, for detailed discussion). The wider field of data collection must include survey sampling. Spatial sampling can be considered as a merging of ideas from survey sampling and optimum design. A recent article which draws together many of the ideas is Ylvisaker (1986). In the present paper a simple criterion and examples are presented in a deliberate attempt to understand the spectrum between controlled experiment and more passive observation. This contrast has been discussed with some philosophical references in Wynn (1982a). The other distinction which we hope to illuminate is that between observation on a finite, closed, system and observation in an infinite or wider environment.