Effect size
|
Shift of median to
|
ICCa
|
SD of random village effect
|
Child in every HH
|
Child in 50% HH
|
---|
Power
|
Power
|
---|
0.27 SD
|
P61
|
0.01
|
0.1 SD
|
99%
|
86%
|
|
0.02
|
0.14 SD
|
98%
|
84%
|
|
0.03
|
0.18 SD
|
97%
|
83%
|
|
0.04
|
0.21 SD
|
96%
|
81%
|
|
0.05
|
0.23 SD
|
95%
|
79%
|
|
0.10
|
0.33 SD
|
86%
|
–
|
0.25 SD
|
P60
|
0.01
|
0.1 SD
|
97%
|
80%
|
|
0.02
|
0.14 SD
|
96%
|
78%
|
|
0.03
|
0.18 SD
|
95%
|
77%
|
|
0.04
|
0.21 SD
|
93%
|
75%
|
|
0.05
|
0.23 SD
|
91%
|
73%
|
|
0.10
|
0.33 SD
|
80%
|
–
|
0.2 SD
|
P58
|
0.01
|
0.1 SD
|
88%
|
–
|
|
0.02
|
0.14 SD
|
84%
|
–
|
|
0.03
|
0.18 SD
|
81%
|
–
|
|
0.04
|
0.21 SD
|
78%
|
–
|
|
0.05
|
0.23 SD
|
75%
|
–
|
- Table 3 gives the achievable power for different effect sizes and intra class correlation coefficients. The expected power is given both under the assumption that a) a child under 5 years old will be found in each household, and b) that a child under 5 years old is found only in every other household. Intra-class coefficients reported in other African contexts range from 0.01 to 0.05
- SD standard deviation of individual growth rates
- aProportion of variance explained by the factor village