Minimum detectable effect size for main outcomes (accounting for sample
design and clustering)
We have conducted power size calculations using different scenarios based on two main outcome variables related to absenteeism and MHM beliefs. For absenteeism, we use the baseline data collected in 2017 for the impact evaluation of the first phase of SWEDD in Mauritania. Based on these data, we constructed a dichotomous variable identifying whether a girl 15-19 years of age was absent from school at least 3 days over the past month. For the power size calculations using this outcome variable, we assume one baseline survey, two follow-up surveys, an intra-group correlation of 0.25, autocorrelation of 0.5 and power of 80%. Calculations assume an ANCOVA specification. For beliefs, we use the midline data collected in 2021 for the impact evaluation of the first phase of SWEDD in Niger. Based on the data, we constructed a dichotomous variables indicating whether an adolescent girl 12-19 years old thinks that when a girl is menstruating, she should not go to school or only go partially. For the power size calculations, we assume one baseline survey, one follow-up survey, an intra-group correlation of 0.15, autocorrelation of 0.5 and power of 80%. Calculations assume an ANCOVA specification.
For our power size calculations, we use as a reference the impact found by Benshaul‐Tolonen et al. (2019) on absenteeism. Authors found that girls that received sanitary pads were 5.4 percentage points less likely to be absent at endline, with a reference value of 14%, which would imply a decrease of 38%. In order to detect a 12pp (percentage points) decrease in absenteeism (i.e., whether a girl 15-19 years of age was absent from school at least 3 days over the past month), we would need 29 secondary schools and 1,516 adolescent girls per arm (accounting for 15% attrition and 80% take-up). To detect a 15pp decrease in absenteeism, we would need a sample of 19 secondary schools and 993 adolescent girls, per arm. In terms of the outcomes related to MHM beliefs, to detect a decrease of 15pp in the likelihood that a girl thinks that when menstruating, a girl should not go or should only go partially to school, we need a sample of 24 secondary schools and 1,255 adolescent per arm (accounting for a 15% attrition and 80% take-up).
These results show that with a sample of approximately 28 secondary schools and 1,598 girls per treatment arm, we will be able to detect a decrease in absenteeism of 12pp to 15pp, and a decrease in misbeliefs about menstruation of 15pp.