Skip to main content
. 2013 Nov 28;2:107. doi: 10.1186/2046-4053-2-107

Table 1.

Power to determine the number of studies needed to show a moderator effect in a given meta-analysis dataset

  Proportion of studies (%) with positive expression of moderator (for example, ‘high quality’)
Studies ( n )
 0.5
 0.3
 0.2
 0.1
Moderator effect = 0
200
 6
 6
 5
 6
400
 6
 5
 6
 5
600
 5
 6
 6
 6
800
 6
 5
 6
 7
Moderator effect = 0.1
200
 24
 22
 22
 13
400
 45
 40
 33
 21
600
 65
 54
 48
 29
800
 75
 69
 53
 39
Moderator effect = 0.2
200
 75
 67
 57
 32
400
 95
 93
 84
 60
600
 100
 99
 96
 78
800 100 100 98 90

Extrapolated from random sample of 200 RCTs, I2 = 92%, τ2 = 0.285; mean sample size 132; alpha = 0.05; in the absence of a moderator effect (moderator effect = 0), the power should vary around 5%.