论文部分内容阅读
一在经典测量理论中,人们已经认识到,测验结果的信度会受到考生分数分布范围的影响。考生分数的分布范围越大,信度系数也越大。但人们在阐述这个问题时,一般都是以相关系数为例来说明的,因为相关系数会受到变量取值范围的影响,而有些信度系数也正是用相关系数来表示的,例如:重测信度、复本信度和分半信度,因此,认为这些信度系数将受到分数分布范围的影响,这显然是合理的。但是,目前我国的许多教育测量工作者,还经常采用α系数表示测验结果的信度,α系数是一
In the classical measurement theory, it has been recognized that the reliability of the test results will be affected by the range of candidates score distribution. The larger the distribution range of the candidate scores, the greater the reliability coefficient. However, when people describe this issue, they generally use the correlation coefficient as an example to explain, because the correlation coefficient will be affected by the variable range of values, and some reliability coefficient is also expressed by the correlation coefficient, for example: heavy Test reliability, replica reliability and sub-half reliability, so that these reliability coefficients will be affected by the scope of the fractional distribution, which is obviously reasonable. However, at present, many education surveyors in our country often use the alpha coefficient to express the reliability of the test result. The alpha coefficient is one