Table 2 Minimum detectable differences for sample estimates

Kibera

600 girls per arm at follow-up (2019)

Percent of girls who have given birth: Assuming that 15.4 % of girls in the violence prevention only arm would have given birth by follow-up, can detect a statistically significant difference of 5.4 percentage points between the violence prevention only arm and each of the other three arms

Grades of schooling: Assuming a correlation coefficient of 0.33, can detect a statistically significant difference of 0.49 grades of schooling between any two arms

480 girls per arm at follow-up (2019)

Assuming that 15.4 % of girls in the violence prevention only arm would have given birth by follow-up, can detect a statistically significant difference of 6.3 percentage points between the violence prevention only arm and each of the other three arms

Assuming a correlation coefficient of 0.33, can detect a statistically significant difference of 0.55 grades of schooling between any two arms

Wajir

20 clusters per arm, 32 girls at follow-up (2019)

Assuming that 17.6 % of girls in the violence prevention only arm would have given birth by follow-up, can detect a statistically significant difference of 5.9 percentage points between the violence prevention only arm and each of the other three arms

Assuming a correlation coefficient of 0.26, can detect a statistically significant difference of 0.48 grades of schooling between any two arms

20 clusters per arm, 22 girls at follow-up (2019)

Assuming that 17.6 % of girls in the violence prevention only arm would have given birth by follow-up, can detect a statistically significant difference of 6.9 percentage points between the violence prevention only arm and each of the other three arms

Assuming a correlation coefficient of 0.26, can detect a statistically significant difference of 0.49 grades of schooling between any two arms