COMPARING INDIVIDUAL MEANS IN THE ANALYSIS OF VARIANCE
COMPARING INDIVIDUAL MEANS IN THE ANALYSIS OF VARIANCE
复制标题
DOI：
10.2307/3001913
复制
发表时间：
1949-01-01
期刊：
影响因子：
1.9
通讯作者：
TUKEY, JW
中科院分区：
文献类型：
--
作者：
TUKEY, JW
例如，7个变量。荞麦A、B、C、D、E、F和G在12个地点的每一个都进行了产量测试，并且希望得出“B和F产量优于A、C和G，而A、C和G产量优于D和E”类型的结论，将有120，904个这样的陈述，每个陈述都需要显著性测试。这个明显不切实际的程序可以被另一个基于3个标准的差异。这意味着：（1）变量之间的差距过大。（2）当一个变量被设置为一个变量时，平均值偏离大平均值太多，以及（3）var.手段加在一起变化太大。该程序被系统化为6个简单的步骤。接下来是对distr的讨论。的max。gap、学生化极端偏差、distr.长间隙的对称情况和实验结果。取样.
If for example, 7 vars. of buckwheat, A, B, C, D, E, F, and G, have been tested for yield in each of 12 locations, and it is desired to draw a conclusion of the type "B and F yield better than A, C, and G, which yield better than D and E," there will be 120,904 such statements, each requiring tests of significance. This obviously impracticable procedure can be replaced by another based on 3 criteria for differences between var. means: (1) there is an unduly wide gap between var. means when arranged in order of size, (2) one var. mean straggles too much from the grand mean, and (3) the var. means taken together are too variable. The procedure is systematized into 6 simple steps. This is followed by discussions of the distr. of the max. gap, the studentized extreme deviate, the distr. of long gaps, the symmetrical case and results of exptl. sampling.