Designing experiments and analyzing data

Abstract

Appropriate experimental design and analysis of data are fundamental to furthering our understanding of the impact of trees on soil fertility. Although there are many standard statistical texts that describe procedures for agronomic or forestry trials, the inclusion of one or more tree components in an agricultural context tends to increase both the spatial variability and the time horizon that have to be considered. This chapter makes reference to places where standard designs and analyses are described, but focuses on aspects of particular importance in an agroforestry context. Given the long-term nature of many agroforestry practices, setting appropriate objectives for individual experiments is of vital importance but is seldom discussed in standard texts and is often the root cause of failure to achieve useful outcomes from agroforestry trials. Here, advice on setting objectives is given before consideration of appropriate treatment structures, layout and replication. Experiments with tree fallows are given special attention. They have become particularly important in soil fertility research over the last decade and require careful consideration because they generally involve a large number of factors, are subject to interference among plots and include treatments with different fallow lengths and hence timing and duration. The spatial and temporal heterogeneity in soils induced by the presence of a relatively few, large plants that grow over several years complicate sampling strategies, so advice is given on the most efficient sampling schemes to use to meet different types of experimental objectives