Friday, January 11, 2013

LANGUAGE ASSESSMENT: TEST RELIABILITY

  -->
Characteristics of a Good Test: Reliability

          A measure of the consistency with which the question, test or examination produces the same results under different, but comparable conditions. A reliable assessment item gives reproducible scores with similar populations of students, and is therefore as independent of the characteristics and vagaries of individual markers as possible. This is often difficult to achieve in practice. That is called reliability test.
          It is obviously important to have reasonably reliable assessment procedures when a large number of individual markers assess the same question (e.g in national school examinations or with many postgraduates marking lab work). A student answer which receives a score of 75 per cent from one marker and 35 per cent from another, for example, reveals a patently unreliable assessment procedure.
          To help produce reliability, the questions which comprise an assessment should (ideally) test only one thing at a time and give the candidates no choice. The assessment should also adequately reflect the desired outcomes of the teaching unit. Note that the reliability and validity factors in an assessment are in no way directly linked - a test or examination, for example, may be totally reliable and yet have very low validity, and vice versa.
.
-->


          No test will be absolutely consistent. Even physical measurements such as height and weight will vary from measurement to measurement, and psychological measurements are bound to be even less consistent. Nonetheless, it is necessary to make psychological tests as reliable as possible.
          There are several ways to measure the reliability of a test. The first method is called test-retest reliability, and it involves giving the same test to the same people at different times. The closer the scores are for the individual at the different times, the more reliable the test is. The problem with test-retest reliability is that people may remember the test from taking it earlier and therefore give the same answers. This may inflate the reliability scores of the test.
          Another method of testing reliability is called split-half reliability. The measurement of reliability compares the scores on two halves of the same test to measure reliability.
          Yet another method of measuring the reliability of a test is called equivalent form reliability. In this method, two different forms of a test are administered to the same people, and the correlation between the scores on the two tests is calculated.
. -->
http://academics.tjhsst.edu/psych/oldPsych/ch11/11.htm

No comments:

Post a Comment