Factors affecting power: Variance (1 of 2)
The larger the
variance
(σ²), the lower the power. In the
formula for z:
increasing σ² increases the denominator
and therefore lowers z and power. For
the example,
σ is the standard deviation of the difference
scores. The power of the test using the .05
significance
level, for N = 25, µ
diff= 20, and various values
of σ is shown in the table on the right side
of this page. The table below shows that the power decreases as σ
increases.
σ |
Power |
50
75
100
125
150 |
0.52
0.26
0.17
0.12
0.10
|
There are ways that an experimenter can reduce variance to
increase power. One is to define a relatively homogeneous population.
For instance, if one were studying reading speed, one could begin by studying
the population of people in their first year at a selective college rather
than the population of all English-speaking adults. The variance would
be much reduced.