On "Path analysis in genetic epidemiology: a critique".
On "Path analysis in genetic epidemiology: a critique".
复制标题
DOI：
--
复制
发表时间：
1983-07
影响因子：
9.8
通讯作者：
Sewall Wright
中科院分区：
文献类型：
--
作者：
Sewall Wright
虽然Karlin et al.[1]的批评主要关注最近由Rao和Morton[2,3]、Rice et al.[1]和Cloninger[5]将通径分析应用于遗传流行病学，但它在通径分析的普遍缺陷上投入了大量的空间。由于后两位作者发表了一篇为他们的方法辩护的文章，我将把自己主要局限于一般的路径分析辩护，这是我在1918年提出的一种方法，在1921年给出了第一个一般性的说明。这种方法的目的是评估某一特定种群中引起变异的各种原因的相对重要性。应该强调的是，它只涉及由于原因而引起的急剧变化，而不是绝对意义上的原因。等基因群体中性状的发展完全依赖于适当的遗传和适当的环境。两者都是绝对必要的，不可能对它们的相对重要性进行评估。然而，观察到的变异必须100%归因于环境变异，因为根据“等基因”的定义，没有任何变异可以归因于遗传变异（除了非常罕见的突变）。W. E. Castle教授是我的研究生导师，他参与了一场争论，争论的问题是，一个群体中总尺寸的差异主要是由于影响身体所有部位生长的因素造成的，还是仅仅是由于不同部位的总和效应造成的。他让他的研究生h·d·菲什（h·d·Fish）和我计算一个以前的学生在一个相当异质的兔子种群中进行的5次骨骼测量之间的所有10种相关性。这10个系数都比较高。558到0.758)。卡斯尔教授以此来支持他的观点，反对某些人的观点，即种族杂交可能导致后代的不和谐发展。然而，这种相关性远非完美。试图对影响总体尺寸的因素、影响某些测量组的因素和影响限于特定测量的因素的相对贡献——对其平方标准差——做出更精确的评估，似乎很有趣。这种方法后来被称为路径分析，只不过系数是路径系数的平方。
While the critique by Karlin et al. [1] was primarily concerned with recent applications of path analysis to genetic epidemiology by Rao and Morton [2, 3], Rice et al. [4], and Cloninger [5], it devoted much space to alleged shortcomings of path analysis in general. Since the latter authors are publishing a defense of their methods [6], I will confine myself largely to the defense of path analysis in general, a method that I proposed in 1918 [7], with first general account given in 1921 [8]. The purpose of this method is the evaluation of the relative importance of the various causes of variation in a particular population. It should be emphasized that it was concerned only with the precipitating variations due to causes, not with causes in an absolute sense. The development of characters in a population that has been made isogenic are absolutely dependent on an appropriate heredity and also on an appropriate environment. Both are absolutely necessary, and no evaluation of their relative importance is possible. Observed variation, however, must be due 100% to environmental variation, since none can be due to genetic variation by definition of "isogenic" (apart from very rare mutations). Prof. W. E. Castle, my mentor as a graduate student, became involved in a controversy on whether differences in total size in a population are due predominantly to factors that affect the growth of all parts of the body alike or are merely due to the summation effect, on various parts separately. He asked his graduate students, H. D. Fish, and myself to calculate all of the 10 correlations among five bone measurements made by a former student in a rather heterogeneous population of rabbits. The 10 coefficients were all rather high (.558 to .758). Prof. Castle took this as supporting his position [9] in opposition to the view of certain others that racial crossing was likely to lead to disharmonious development in later generations. The correlations were, however, far from perfect. It seemed of interest to attempt to make a more precise evaluation of the relative contributions of factors affecting general size, of ones affecting certain groups of measurements, and of ones with effects restricted to the particular measurement: to its squared standard deviations. The method arrived at was what was later called path analysis, except that the coefficients were the squares of the path coefficients.