Experimental power

Explanations > Social Research > Statistical principles > Experimental power

Description | Example | Discussion | See also

Description

The power of a test is the probability that it will detect an effect (assuming one exists). A 'powerful' test is thus more useful and preferable to an experiment that is less powerful.

A Type 2 error is one where an effect that exists is not found. The probability of this is beta (b). The power is thus the ability of an experiment to avoid type 2 errors and can be calculated as:

Power = (1 - b)

Example

The probability of a a student who has passed a school test actually being incapable is calculated as 20%. The Power of the test, the chance of a passing student being capable, is thus 80%.

Discussion

It is possible to identify the size of the sample needed to reach a given power for any test, although this requires complex calculations. Statistical tools may do this for you.