|
Field
Power calculation: Minimum Detectable Effect Size for Main Outcomes
|
Before
For each of the main outcome variables considered, we determined the power of our design by employing a J-PAL algorithm for a stratified research design . We obtained the required parameter estimates for each of the outcome variable from our baseline survey of 472 students. Parameters retrieved from the baseline survey data for a particular outcome variable include the intra-cluster correlation coefficient, within-cluster variation, and the control-group intercept. We provide power estimates for each of our main outcome variables that were described in section 2.2. Our sample size of 472 individuals stratified across three madrasahs yield the following power estimates:
• Depression index: power of 0.331 for a minimum detectable effect of 10 percent over the average amount.
• Growth index: yields a power of 0.32 for a minimum detectable effect of 10 percent over the average amount
• School anxiety: yields a power of 0.367 for a minimum detectable effect of 10 percent over the average amount
• Anger inventory: yields a power of 0.37 for a minimum detectable effect of 10 percent over the average amount
• Future outlook: yields a power of 0.33 for a minimum detectable effect of 10 percent over the average amount
• Self-perception: yields a power of 0.392 for a minimum detectable effect of 10 percent over the average amount
• Peer relations : yields a power of 0.33 for a minimum detectable effect of 10 percent over the average amount
• Perceived safety and abuse: yields a power of 0.31 for a minimum detectable effect of 10 percent over the average amount
|
After
For each of the main outcome variables considered, we determined the power of our design by employing a J-PAL algorithm for a stratified research design . We obtained the required parameter estimates for each of the outcome variable from our baseline survey of 472 students. Parameters retrieved from the baseline survey data for a particular outcome variable include the intra-cluster correlation coefficient, within-cluster variation, and the control-group intercept. We provide power estimates for each of our main outcome variables that were described in section 2.2. Our sample size of 472 individuals stratified across three madrasahs yield the following power estimates:
• Growth index: yields a power of 0.931 for a minimum detectable effect of 10 percent over the average amount
• Future outlook: yields a power of 0.915 for a minimum detectable effect of 10 percent over the average amount
• School anxiety: yields a power of 0.81 for a minimum detectable effect of 10 percent over the average amount
• Anger inventory: yields a power of 0.76 for a minimum detectable effect of 10 percent over the average amount
• Depression index: power of 0.668 for a minimum detectable effect of 10 percent over the average amount.
• Self-perception: yields a power of 0.893 for a minimum detectable effect of 10 percent over the average amount
• Peer relations : yields a power of 0.84 for a minimum detectable effect of 10 percent over the average amount
• Perceived safety and abuse: yields a power of 0.91 for a minimum detectable effect of 10 percent over the average amount
|