Two factors suggest that experimental designs used in on-station research are not appropriate for on-farm agroforestry research. First, successful technology development and validation requires farmer participation, which in turn requires more flexible experimental designs for on-farm experimentation. Second, the advent of widely available statistical computer packages and computing power allows the experimenter to deviate from ‘standard’ design restrictions of complete blocks and full replication. Farmer involvement in the research process should include the opportunity for farmers to ask questions and to define treatments along-side those of the researcher. To make inferences about farmer-defined augmented treatments that may only appear on one farm requires estimates of farm-by-treatment effects. This estimate can be obtained from the associated researcher's treatments that are applied to all of the farms participating in the trial. The use of augmented designs minimises plot number, while still enabling the researcher's and farmer's questions to be answered. The proposed design fills a methodological gap between informal farmer observation trials (where no statistical analysis is attempted) and larger scale extension trials (composed of simple treatment comparisons tested over a large number of farms).
DOI:
https://doi.org/10.1007/BF00120192Altmetric score:
Dimensions Citation Count:
Publication year
1991
Authors
Language
English
Keywords
agroforestry, farmers, on-farm research, research
Geographic
Kenya
