Estimates of stability for comparing varieties

Abstract

For purposes of comparison of analytical methods the yield data on pearl millet hybrids and varieties for five years from locations in India and Pakistan have been analysed using a regression analysis and a meanstandard deviation analysis. The results of the mean-standard deviation analysis and the regression analysis were similar whether carried out on all environments, or on the highest- and lowest-yielding sets of environments. This was substantiated by the remarkable correlation between the slopes from the regression analysis and the standard deviations whatever environmental set was considered. The validity of using a single year's across location data with the mean-standard deviation analysis, if choice-theoretic criteria are used, was examined. It is concluded that, although further confirmation is required, single-year data seem to suffice. The relative merits of the mean-standard deviation analysis and a regression analysis are discussed; the choice-theoretic framework of the mean-standard deviation analysis is advantageous and complementary to the regression analysis