Minimum detectable effect size for main outcomes (accounting for sample
design and clustering)
The average sales revenue was MWK 52,500, and the standard deviation was MWK 75,000. A third of the sample farming households did not engage in crop sales. For those who did engage in crop sales, the average sales amount was MWK 83,450, and the standard deviation is MWK 80,750. With the list of 118 producer groups identified by the project that they will support (both batches: treatment and control groups), and 9 members to be randomly selected per group, and the computed intraclass correlation coefficient (ICC) is 0.17, the target sample size can detect a minimum effect of 38%. The project targets a 30% increase in average sales revenue among selling households during baseline and, for those not selling during baseline, to sell at an average of MWK52,500 similar to the baseline sellers. These represent a 40% increase over the average sales revenue among the baseline sellers (both represent an additional average sales revenue of MWK 44,100). The target sample size is just about what the project is targeting is a small allowance for potential attrition.