The relationship of validity coefficients to the practical effectiveness of tests in selection: discussion and tables.
The relationship of validity coefficients to the practical effectiveness of tests in selection: discussion and tables.
复制标题
有效性系数与选择测试的实际有效性的关系：讨论和表格。
DOI：
--
复制
发表时间：
1939
期刊：
影响因子：
--
通讯作者：
J. T. Russell
中科院分区：
文献类型：
--
作者：
H. Taylor;J. T. Russell
经常有人指出，相关系数的大小本身并不能充分代表所考虑的两个变量之间关系的大小。许多人提出了许多不同的统计常数，以更满意地表示两个变量之间关系的真实的大小。所有这些常数都是皮尔逊r的相对简单的函数。这些包括一类函数的r直接承担的程度，其中一个变量可以预测的其他评估时，相关系数是一个给定的幅度。r的这些函数包括凯利的异化系数k，它是\A~**! ^这样的函数k作为船体的措施的效率，E，这是1-k，和奥德尔的g，这是k/\/2。所有这些评估相关系数的方法都有一个共同的基本特征，这可以从它们都与异化系数密切相关的事实中看出。也就是说，随着相关系数的大小增加，一个变量可以从另一个变量预测的程度越来越快。例如，0.50的r通常被认为只有1.00的r的13%好。据认为，0.87的r值仅为1.00的r值的一半。广泛接受这些措施作为正确的方法来评估相关系数带来了一个
IT has often been pointed out that the magnitude of the correlation coefficient as such is not an adequate representation of the magnitude of the relationship between the two variables which are under consideration. A number of different statistical constants have been proposed by various persons, as giving a more satisfactory representation of the real magnitude of a relationship between two variables. All of these constants are relatively simple functions of the Pearson r. These include a class of functions of r which bear directly upon an evaluation of the extent to which one variable may be predicted from the other when the correlation coefficient is of a given magnitude. These functions of r include Kelly's alienation coefficient, k, which is \ A ~ **! ^such functions of k as Hull's measure of efficiency, E, which is 1-k, and Odell's g, which is k/\/2. All of these ways of evaluating the correlation coefficient have one fundamental characteristic in common, as might be expected from the fact that they are all closely related to the alienation coefficient. That is, that as the size of the correlation coefficient increases the extent to which one variable can be predicted from the other increases more and more rapidly. For example, an r of .50 is usually considered to be only 13% as good as an r of 1.00. It is considered that an r of .87 is only half as good as an r of 1.00. The wide-spread acceptance of such measures as the correct way of evaluating correlation coefficients has brought about a