懿说学区（47） SPSS统计分析（57）内在信度分析

Yi Shuo School District (47) SPSS Statistical Analysis (57) Internal Reliability Analysis

【思维导图】

“分享兴趣，传播快乐，增长见闻，留下美好！大家好，这里是小编。欢迎大家继续访问学苑内容，我们将竭诚为您带来更多更好的内容分享。

"Share interests, spread happiness, increase knowledge and leave beauty! Hello everyone, this is Xiaobian. Welcome to continue to visit the content of the academy, and we will bring you more and better content sharing wholeheartedly.

【基本概念】

内在信度也称为内部一致性，用以衡量组成量表题项的内在一致性程度如何。常用的检测方法是Cronbachα系数法和折半（Split-half）系数法。

Internal reliability, also known as internal consistency, is used to measure the internal consistency of the items in the scale. The commonly used detection methods are Cronbach α coefficient method and Split-half coefficient method.

其中，Cronbach α系数是最常用的信度系数。其信度系数在0~1之间。根据经验来看，如果α大于等于0.9，则认为量表的内在性都很高；如果α大于等于0.8、小于0.9，则认为内在性比较好；如果α大于等于0.7、小于0.8，则认为量表设计可以接受；如果α大于等于0.6、小于0.7，则认为量表设计勉强可以接受，最好进行适当修改；如果α大于等于0.5、小于0.6，表明量表设计不理想，需重新编制或修订；如果α小于0.5表明量表非常不理想，应舍弃。

Among them, Cronbach α coefficient is the most commonly used reliability coefficient. Its reliability coefficient is between 0 and 1. According to experience, if α is greater than or equal to 0.9, it is considered that the internality of the scale is very high; If α is greater than or equal to 0.8 and less than 0.9, it is considered that internality is better; If α is greater than or equal to 0.7 and less than 0.8, the scale design is considered acceptable; If α is greater than or equal to 0.6 and less than 0.7, it is considered that the scale design is barely acceptable and it is best to modify it appropriately; If α is greater than or equal to 0.5 and less than 0.6, it indicates that the scale design is not ideal and needs to be re-compiled or revised; If α is less than 0.5, it indicates that the scale is very unsatisfactory and should be discarded.

折半信度是在测试后对测试项目按奇项、偶项或其他标准分成两半，分别记分，由两半分数之间的相关系数得到信度系数，因此它实际上是检验一个测试内部一致性的粗略估计，折半信度是建立在相关系数基础上的，但它在相关系数（由于折半，所以相关系数只是半个测验的信度，可能会低估原测度的信度）基础上，需要进行斯皮尔曼-布朗公式校正。

Half-cut reliability is to divide the test items into two halves according to odd items, even items or other standards after testing, and score them separately. Therefore, it is actually a rough estimate to test the internal consistency of a test. Half-cut reliability is based on the correlation coefficient, but it needs to be corrected by Spielman-Brown formula on the basis of the correlation coefficient (because of half-cut, the correlation coefficient is only the reliability of half test, which may underestimate the reliability of the original measure).

折半信度面临的主要问题是如何将问题分成两半，一般事实式的问题是不太容易折半的，比如年龄和教育程度是无法相比的。因此这种方法一般不适合于事实式量表。态度式量表一般围绕某个主题进行多种正、反方面的陈述，由被调查者对陈述进行选择，例如“很不满意”，“不满意”，“既非满意也非不满意”，“满意”，“很满意”中的一个，对以上5种选择，分别赋予1~5分，然后将该量表的全部题项分成尽可能相近的两半，按前后两部分或按题号的奇偶性分都可以，只是要注意两部分必须尽可能相当（内容、形式、题数等），不同的折半法可能会得到不同的结果。

The main problem of halving reliability is how to divide the problem into two halves. General factual problems are not easy to be halved, for example, age and education level cannot be compared. Therefore, this method is generally not suitable for factual scale. Attitude scale generally makes a variety of positive and negative statements around a certain topic, and the respondents choose the statements, such as "very dissatisfied", "dissatisfied", "neither satisfied nor dissatisfied", "satisfied" and "very satisfied". For the above five choices, 1 ~ 5 points are given respectively, and then all the items of the scale are divided into two halves as close as possible, which can be divided according to the parity of the two parts or the number of questions, but it is important to pay attention to the fact that the two parts must be as close as possible (content, form, number of questions, etc.)

【案例分析】

接下来我们通过一个案例来深入了解内在信度分析。

Next, we use a case to deeply understand the internal reliability analysis.

在学生的性格特征调查中，共选了10名学生在8个项目上进行测试，其数据如下表所示，试对其进行内在信度分析。

In the investigation of students' personality characteristics, 10 students were selected to test on 8 items, and the data are shown in the following table, so as to try to analyze their internal reliability.

第1步，分析，本例通过Cronbachα系数来衡量其内在一致性。

In step 1, analysis, this example measures its internal consistency by Cronbach α coefficient.

第2步，数据组织，按如表所示的表头定义变量，输入数据并保存。

Step 2, data organization, define variables according to the header shown in the table, input data and save it.

第3步，按如下图所示，进行内在一致性分析的设置。

Step 3, as shown in the following figure, set up the internal consistency analysis.

第4步，主要结果及分析。

Step 4, main results and analysis.

折半系数法：对学生的性格特征数据用折半法，求信度系数。

Half-fold coefficient method: The reliability coefficient is calculated by half-fold method for students' personality data.

具体过程和设置与上面相似，主要结果如下所示。

The specific process and setup are similar to the above, and the main results are as follows.

下期预告：本期，我们学习了信度分析的理论基础。下一期，我们将会学习内在信度信度分析的基础知识。

In this issue, we learned the theoretical basis of reliability analysis. In the next issue, we will learn the basic knowledge of internal reliability analysis.

如果您对今天的文章有独特的想法，欢迎给我们留言，让我们相约明天，祝您今天过得开心快乐！

If you have a unique idea of today's article, welcome to leave us a message, let us meet tomorrow, I wish you a happy today!

参考资料：《SPSS23（中文版）统计分析实用教程》、百度百科

翻译：讯飞语音