Replica reliability, two parallel tests are compiled to test the same group of people, such as split-half reliability. After testing the same group of people, the test results are divided into two to find correlation. Finally, Kolumbach coefficient is a general method to find reliability, but in fact it is not a reliability test, but depends on the internal consistency of the scale.

If you look at the correlation between all items, if the correlation is high and the α coefficient is high, you will think that the problems in the test are homogeneous and similar. Then, if the score of the test is half, it should be highly correlated, and it is approximately considered as good reliability.

Extended data:

Crombach put forward that it is not required that the test item has only (0, 1) points, and it can handle the calculation of internal consistency coefficient of any test.

Generally speaking, the higher the coefficient, the higher the reliability of the tool. In basic research, the reliability must reach at least 0.80 to be accepted. In exploratory research, as long as it reaches 0.70, it is acceptable, and the reliability between 0.70 and 0.98 is high, and the reliability below 0.35 is low, so it must be rejected.

Kronbach coefficient formula? α? ﹦(n？ /? n？ - 1)( 1-∑Si2/St2).

α? Is the reliability factor, n? In order to test the number of questions, Si2 represents the variance of all subjects' scores on question I, and St2 represents the variance of all subjects' total scores.

The α coefficient is the average value of all possible split-half reliability, and it is only an estimate of the lower bound of test reliability, that is, a large α value must have a high reliability, while a small α value cannot determine a low reliability.

Baidu Encyclopedia-Kronbach Coefficient