Secondly, there are any number of ways in which reliability measures can be obtained, and the magnitude of the reliability coefficient will be a direct reflection of the particular approach used. Some broad definitions are described below:
1. Internal consistency. Measures of internal consistency arc based on a single administration of the measure. If the measure has a relatively large number of items addressing the same underlying dimension; for example, ‘Are you able to dress yourself?’, ‘Are you able to shop for groceries?’, ‘Can you do the sewing?’ as measures of physical function, then it is reasonable to expect that scores on each item would be correlated with scores on all other items. This is the idea behind measures of internal consistency—essentially, they represent the average of the correlations among all the items in the measure. There are a number ways to calculate these correlations, called Cronbach's alpha, Kuder--Richardson, or split halves, but all yield similar results. Since the method involves only a single administration of the test, such coefficients are easy to obtain. However they do not take into account any variation from day to day or from observer to observer, and thus lead to an optimistic interpretation of the true reliability of the test.
Secondly, there are any number of ways in which reliability measures can be obtained, and the magnitude of the reliability coefficient will be a direct reflection of the particular approach used. Some broad definitions are described below: 1. Internal consistency. Measures of internal consistency arc based on a single administration of the measure. If the measure has a relatively large number of items addressing the same underlying dimension; for example, ‘Are you able to dress yourself?’, ‘Are you able to shop for groceries?’, ‘Can you do the sewing?’ as measures of physical function, then it is reasonable to expect that scores on each item would be correlated with scores on all other items. This is the idea behind measures of internal consistency—essentially, they represent the average of the correlations among all the items in the measure. There are a number ways to calculate these correlations, called Cronbach's alpha, Kuder--Richardson, or split halves, but all yield similar results. Since the method involves only a single administration of the test, such coefficients are easy to obtain. However they do not take into account any variation from day to day or from observer to observer, and thus lead to an optimistic interpretation of the true reliability of the test.
การแปล กรุณารอสักครู่..
