Reliability
Posted by David Richards on August 23, 2011
Reliability refers to the ability of the same measurement to produce consistent results over time. Some data is inherently reliable (such as a person’s age) and other more subjective data is not very reliable.
Another way of putting it is that reliable tests are free from significant test errors.
Reliability can be interpreted as an index of test results error. Generally, tests with reliability coefficients at or above 75 percent (p=0.75) are considered reliable.
If we were to throw darts at a board, reliability is that the darts are grouped together closely, and not randomly placed on across the board. Validity is that the darts are at the center of the target. Furthering this example:
- If the darts are clumped together but not in the center, that is reliable, but not valid.
- If the darts are clumped together (reliable) and in the center, that is both reliable and valid.
Source: ASTD’s Measuring and Evaluating, pp. 6-7
