Statistical analysis of on-farm yield trials for testing adaptability of rice

Abstract

Selection of rice varieties with wide adaptability over diverse farming environments is important, prior to varietal recommendation in order to achieve a high rate of varietal adoption particularly in developing countries. Multi-locational rice varietal adaptability trials conducted in farmers’ fields consist of only few varieties and minimum replicates within a location. The statistical methodology available at present for analysis of these trials is inadequate. This paper proposes an efficient and simple method to analyse on-farm multi-locational yield trials with few varieties and minimum replicates, for evaluating wide adaptability and stability of rice over diverse environments. In the present method of analysis, deviations of the plot yields from the maximum plot yield in a given environment are computed and a pooled analysis of variance is run on these deviations. Mean deviation (D) and variance in deviations (v²) across locations for each variety are computed. Varieties are evaluated for general adaptability using D and v², the stability parameter. Variety with the lowest D and non-significant v² is selected as the most adaptable variety in the test since it is the one with the highest yield in all environments. The present method is illustrated in comparison with the most appropriate reference method that provides a parameter called ‘superiority measure of varietal performance’ on on-farm yield trials in rice. While retaining the advantages of the reference method, the present method has an additional advantage of efficient and precise detection of varietal differences in their adaptability over diverse environments in an objective and realistic manner.