Comparisons of the Performance of Sugar-cane Varieties in Trials and in Agriculture

Abstract

SUMMARYTwelve varieties of sugar-cane grown in Barbados, West Indies, during the past 40 years are compared with respect to cane yield performance in trials and in agriculture. Comparisons are based on data-pairs matched for site, season and crop, using both means and variety-on-variety regressions. Conclusions are that: trials greatly over-estimate mean agricultural differences between varieties; variety-on-variety regressions reveal numerous GE effects and account for much variance, but agree poorly as between trials and agriculture and are not useful in improving prediction; in view of the universal use of trials for prediction and decision-making, it is a matter of some practical importance that their validity should be tested in other crops.