Who knows the reliability and validity of the questionnaire?

Test reliability, also known as test reliability, refers to whether the test results are stable and reliable. In other words, does the test result reflect the actual language level of the subjects? For example, if the same set of tests scores high or low in several tests on the same test subject (that is, the subjects themselves have not changed), it means that the test lacks reliability. The reliability of the test is closely related to its validity. Generally speaking, only tests with higher reliability have higher validity, but higher validity does not guarantee higher reliability. The reliability of the test mainly involves the reliability of the test questions themselves and the reliability of the scores. Whether the test questions are reliable depends mainly on the scope, quantity and discrimination of the test questions. Whether the score is reliable depends on whether the scoring standard is objective and accurate.

The reliability of the test is usually expressed by the correlation coefficient (that is, the proportional relationship between two numbers). The greater the correlation coefficient, the higher the reliability. When the coefficient is 1.00, the reliability of the test reaches the highest level; When the coefficient is 0.00, the reliability of the test is minimum. Under normal circumstances, the coefficient will not be as high as 1.00, nor will it drop to 0.00, but it will be somewhere in between. The requirements for reliability indicators vary according to different test types. People usually require the reliability coefficient of standardized tests to be above 0.90. For example, the reliability of TOEFL is about 0.95, while the reliability coefficient of classroom test is between 0.70 and 0.80. There are many ways to calculate the reliability of the test. Here are only three easy-to-operate methods:

1) retest method. Using the same set of papers to test the same group of subjects at two different times, so as to get two groups of scores, and then calculate the correlation coefficient of the two groups of scores. Of course, in the two tests, students' scores in the second test should be higher than those in the first test, because students have made progress in the second test and gained more field experience. But if the question is reliable, the ranking order of each student in the two tests should remain basically the same.

2) Alternative methods. Two groups of equivalent test papers with the same test type and the same difficulty, but different specific topics, are tested twice, and then the correlation coefficient of the two scores is calculated.

3) Half method. The test is only conducted once, but the whole test paper is divided into two groups according to the odd and even numbers, and the correlation coefficient of the scores of the two groups is calculated, and then the reliability coefficient of the whole test paper is calculated by Spearman-Brown formula. The specific calculation steps are: multiply the correlation coefficient of two groups of scores by 2, then divide by 1, and add the correlation coefficient of two groups of scores.

/view/698674.htm

Test validity, also known as test validity, refers to the degree to which a set of tests measures what should be tested. In other words, whether a set of tests has achieved the expected purpose and measured what it wants to measure. For example: "Is photography an art or a science? Discussion .. "This topic is based on the knowledge of photography and its main contents. If it is used to test language ability, it is ineffective. Another example is dictation to measure students' listening ability, and its validity is not ideal, because the written record of oral English not only relates to students' listening ability, but also relates to students' writing speed, spelling ability, grammar knowledge, memory ability and comprehension ability of the full text.

The effectiveness of the test can generally be divided into the following categories:

1) coupon validity period. Refers to the test paper standard that should be reached, that is, whether a set of test questions is appropriate on the surface. For example, if the reading comprehension test includes many dialect words that the subjects have not learned, it can be considered that the test lacks face validity. Face validity is the guarantee factor to test the normal level of the subjects.

2) Content validity. Refers to whether a set of test questions tests what should be tested or whether the content of the test reflects the test requirements, that is, the representativeness and coverage of the test. For example, if a set of pronunciation skills test questions only examines certain skills necessary for pronunciation, such as the pronunciation of a single phoneme, but does not examine the pronunciation of stress, intonation or phonemes in words, then the content validity of the test is very low.

3) Structural validity. Refers to the degree to which the items of a set of test questions reflect the basic aspects of the theory on which the test is based. For example, based on the structuralist language theory, it is believed that systematic language habits are acquired through sentence patterns, so the test questions that emphasize vocabulary and grammatical environment lose their compilation validity.

4) Empirical validity. Empirical validity is a measure of the validity of a test, which is obtained by comparing the test with one or more standard scales. Empirical validity can be divided into two types: one is * * * validity, that is, the coefficient obtained by comparing the result of one test with the result of another effective test with similar time, or comparing it with the teacher's appraisal; The second is predictive validity, that is, the coefficient obtained by comparing the result of a test with the later language ability or the teacher's later evaluation of students.

Generally speaking, when testing the validity of a test paper, we should not only examine the content of the test paper according to the requirements of the syllabus and the theory of conceptual validity, but also adopt a quantitative method to calculate the correlation coefficient, that is, calculate the correlation coefficient between this test paper and another test paper that has been determined to correctly reflect the level of the subjects. The higher the coefficient, the greater the effectiveness. The validity of classroom test should be between 0.4 and 0.7, and the validity of large-scale test should be above 0.7.