Evaluating cultivars in unbalanced on-farm participatory trials.

摘要: Abstract Data from on-farm participatory varietal trials are often highly unbalanced. Using statistically robust techniques we showed that such data amenable to sophisticated analyses. 38 varieties of wheat tested in 44 mother (all test compared) and 663 baby (single entries compared with local checks) over 6 years Lunawada, Gujarat, India, were analysed. Two combined analyses for grain yield performed by using either a fixed effects model within general linear modelling (GLM) or mixed the restricted maximum likelihood (REML) procedure. Both made substantial adjustments raw means estimated mean values two set correlated (R2 = 0.89) as t-value cultivars check variety (R2 = 0.96). The GLM detected 15 REML 12 significantly superior cultivar 11 common. methods similar identifying top-most promising but was more conservative (i.e. subject Type II error) larger standard errors (S.E.s) n > 17. However, both procedures S.E.s increased sharply when number decreased below 17. data-derived scores (DDYS) odds ratio detecting higher variety. perception-derived (PDYS) ratio, however, four (GW 273, GW 503, K 9107, PBW 343) among top eight other methods. Thus qualitative perceptions could provide cheaper alternative expensive measurements error is not serious consequence. practical utility conclusions new confirmed adoption trends varieties. analyses, maturity indices, identified three categories. Most yielding later maturing and, irrespective their recommendation, farmers quickly adopted them advancing sowing date about month. Of identified, only 273 503 have been released Gujarat. This indicated need policy changes formal testing system widely across states define correct domain cultivation.