# Designing on-farm trials: an example with interventions to improve micronutrient status of grain crops

## TL;DR

This paper explores how to design on-farm trials to effectively evaluate interventions that improve the micronutrient content of grain crops.

## Contribution

The study introduces a systematic approach to designing on-farm trials with sufficient statistical power and spatial coverage.

## Key findings

- Sufficient replication at the farm-scale is critical for detecting treatment effects with high statistical power.
- Regional replication at the scale of farm clusters is needed to estimate treatment means with precision.
- Engaging farmers through visits to experimental sites is necessary to ensure spatial coverage within a 5 km radius.

## Abstract

Design of on-farm experiments to achieve particular objectives, including statistical power, precision of estimates of treatment effects and spatial coverage, remains to be systematically studied. We assessed design options for an extended network to evaluate micronutrient-biofortification interventions for cereal crops in Ethiopia. We identified feasible designs to detect plausible treatment effects with power ≥0.8. Sufficient replication at farm-scale (where each farm is a complete block) was critical for this. To estimate the treatment mean with precision requires sufficient regional replication at the scale of farm clusters. With 250 clusters across the region the median distance from a random point in the region to an experimental site exceeded 5 km, so active steps would be needed to engage farmers, by facilitating visits to experimental sites. The approach used here could be applied more generally to design effective and efficient on-farm experimental networks.