The paper describes the agronomic and statistical principles that form the basis for measuring crop yields in on-farm agroforestry studies. Agroforestry systems differ from agricultural systems because of the presence of tree/crop interfaces and the need for large plots, large borders and long-term monitoring. These differences accentuate the variability of crop performance on farms. Crop yield estimation per unit area in any agroforestry system involves essentially i) stratification of the plot into different, clearly distinguishable crop zones such as those under and free from the influence of trees, those on sloping and flat areas, and those on areas affected by pests, ii) drawing representative samples from each stratum, and iii) weighting the sample yields with weights proportional to the stratum area. The tree/crop interface areas may require further stratification and determination of yields of individual crop rows at different distances away from trees based on the nature and extent of tree/crop competition. The precision of yield estimation depends on how well one is able to define the strata as well as the variance of crop yields in space and time in each stratum. Studies that provide this information are urgently needed for developing practical recommendations for crop-yield measurements on farms.
Tag: on-farm trials
Analyzing ranking and rating data from participatory on-farm trials
Responses in participatory on-farm trials are often measured as ratings (scores on an ordered but arbitrary scale) or rankings (respondents are simply asked to order treatments). Usual analysis of variance and linear model based analyses are not appropriate for these data. Alternative analyses based on generalized linear models are described in this paper. These methods can be successfully used when trial designs are irregular—a common characteristic of participatory trials—and when covariates are measured on each plot or farm in order to identify genotype x environment interaction.