第3章_实验设计.ppt

第3章 实验设计 实验设计的一般原则 实验设计的类型 General principles of Experimental Design Randomization Replication Balancing and Blocking Randomization A historical perspective: The notion of randomization was Fisher’s “great contribution to the scientific method”. Yet W. S. Gossett (“Student “), his mentor and friend, and one of the other giants in the history of statistics, never fully accepted Fisher’s arguments in favor of strict randomization. Worse yet, Gossett argued that systematic designs were superior. -----S. H. Hurlbert, Pseudoreplication and the design of ecological field experiments. 1984, Ecological Monographs I strongly disagree with the suggestion(Cox 1958;Cochran and Cox 1957) that the completely randomized design may be most appropriate in “small experiments” ------S.Hurlbert 1984 In the studies interspersion is more important than randomization because it deals with the critical question of how the experimental units should be distributed in space. -----C.Krebs 1999 Hurlbert reported that in two separate surveys 26% and 48% of the ecological papers surveyed showed the statistical error of pseudoreplication. Underwood (1981) found statistical errors in 78% of the papers he surveyed in marine ecology. Clearly we need to improve the statistical design of ecological studies. ------Charles J. Krebs,1999,Ecological methodology. 我们用t检验比较两组无脊椎动物丰富度的均数： 用错误重复抽样方法来研究火烧对土壤无脊椎动物的影响。在每块土地重复抽取了5个土壤样本。 在这个例子中，应用的统计检验方法是正确的。问题是，重复抽取土壤样本不适合检验火灾的影响。它只能衡量同一场火灾造成的影响。两块样地固有的差异也会对无脊椎动物造成影响。通过这个设计，我们无法得出关于火灾影响的任何结论，我们只能得出这两块样地是不同的结论。 如果每种处理只有一个实验对象，特别当实验周期比较短时，我们可以通过不同时间多次重复实验，用实验组与对照组互换的方法，达到重复的目的。 ------Gerry P. Quinn and Michael J. Keough 2002 Experimental Design and Data Analysis for Biologists. University Press, Cambridge 目前有一本不错的参考书，总结了当研究地点只有一个的时候，研究人类活动可能带来的影响时可以采用的方法。这类实验设计通常称为前后控制影响（BACI）设计（Green1979，Stewart-Oaten等1986），它提供了多种建议，包括在不同时间采样，在不同对照区域采样等等。这些设计都存在争议，对它们正反两方面的评论参见Keough和Mapstone（19