Probability as a Basis for Barley Cultivar Selection by Growers

Inadequate interpretation of cultivar test results can lead to poor cultivar selection by growers and loss of farm income. Analysis of cultivar mean yields does not allow comparison of predicted cultivar responses at different yield levels. Traditional tests of significance are inappropriate for nonacademic clients of applied research. This study introduces cumulative probabilities (CP) as an alternative method of expressing predicted cultivar response across the range of possible yields. Cumulative probabilities, computed using the mean and standard deviation of each cultivar in a normal probability density function, express the probability growers have of obtaining or exceeding given yields when planting each cultivar. Cumulative probabilities are compared with mean separation with LSDs for interpretation of results from 39 single-replicate on-farm drill-strip tests of four spring barley (Hordeum vulgare L.) cultivars conducted in 1990 and 1991 in Washington's malting barley zone. Providing growers with cultivar yield performance alone fails to provide an integrated basis for cultivar selection based on both yield and quality criteria upon which farm income is based. Test weight, kernel plumpness, and protein concentration for feed or malting premiums are integrated with grain yields to estimate net returns for each cultivar. In the farm production environment where the quality of cultivar selection by growers depends upon the quality of easily assimilated information, CPs of yields and expected net returns provide informative criteria for cultivar selection. Research QuestionPublic and private sector representatives, called upon to interpret cultivar trial results for growers without explicitly recommending cultivars, typically present cultivar mean yields with least significant differences (LSDs). Such presentations assume that the differences between cultivars remain constant across all possible yield values and that statistical significance influences grower decisions. Effective use of LSDs to properly separate cultivar means requires an understanding of error risk assessment that is beyond many researchers. We contend that the production obligation to plant at least one cultivar makes statistical significance irrelevant to growers, even if growers were able to properly assess and balance error risks. Presentation of mean yields lead growers to believe that the cultivar with the highest mean yield is the best cultivar. This study presents probabilities as an alternative method of interpretation of cultivar performance trials, which allows comparison of expected cultivar performance across all possible yield levels, and that are easily grasped by nonacademic users. Nonyield cultivar traits are commonly presented separately from yield results, leaving growers to weigh the relative value of yield and nonyield performance before selecting a cultivar. The second objective of this study is to show that an integrated economic interpretation of yield and nonyield traits can enhance the usefulness of cultivar trials. Literature SummaryS.G. Carmer has demonstrated the inadequacies of LSDs using traditional values of α = 0.05 or α = 0.01 for selection of cultivar by growers. He suggested a relatively complex system of risk assessment to adjust the α level in LSDs and minimize error risks important to growers. Carmer's approach to improving cultivar selection is appropriate for situations where cultivars are recommended to growers after risk assessment and analysis by researchers. If cultivars are not recommended, then growers, not researchers, select cultivars and statistical significance appears irrelevant to growers. Growers seem more concerned with expected yields of different cultivars under variable climatic conditions than with the statistical significance of observed mean yield differences. Adapted from econometric methods of risk assessment, probability of avoiding disaster has been proposed to breeders for stability analysis but has not been applied to interpretation of cultivar trial results for growers. Expected net returns for barley cultivars with different yields and different end uses is a useful enhancement of cultivar trial results that may be more important than yield comparisons for cultivar selection. Study DescriptionFour barley cultivars were compared in 1990 and 1991 by eastern Washington growers in 39 environments where spring malting barley can be produced. Two feed cultivars, one malting cultivar, and one dual purpose cultivar were planted in single replicate drill strips ≈2000 ft long and 24 ft wide. Planting and harvesting were done by the growers and all production practices were those normally practiced by the collaborating growers. The probability approach was compared with the traditional presentation of mean values with LSDs for cultivar selection by growers. Cumulative probabilities (CP) of obtaining or exceeding yields ranging from 0 to 7000 lb/acre are computed using the mean and standard deviation of each cultivar in a normal probability density function. Applied QuestionsWhat are the advantages of the probability approach for interpretation of cultivar performance results? Probabilities are versatile, intuitively understandable, easy to compute, and more informative than cultivar means for comparing cultivar performance in variable climatic conditions. Given two assumptions for combining data across locations and years, CPs can be used to determine the proportion of growers expected to obtain a given yield with each test cultivar, or the number of years out of 10 that a grower can expect to meet or exceed yield objectives with test cultivars. For cultivar yield comparisons, growers compare the CPs for different cultivars at the expected yield for prevailing field and climatic conditions. Growers who, for whatever reason, expect yields to be above or below mean yields, can select the cultivar expected to excel in those conditions. Breeders and agronomists can also use probabilities to predict expected cultivar response for quality traits at threshold levels other than mean values. Fewer significant differences were detected with the probability approach than with traditional ANOVA with LSDs. Nevertheless, probabilities are conceptually useful for conveying both the predictability and uncertainty associated with the application of research results to complex biological systems in variable environments. Was the economic interpretation useful? Economic interpretation of trial results provides more easily assimilated information to growers for cultivar selection than interpretation of yield results alone. Steptoe was higher yielding than the other cultivars but three of the four test cultivars, including Steptoe, have nearly identical expected net returns. Economic interpretation encourages agronomists and breeders to remain aware of the relative importance of nonyield traits for the income of grower clients. RecommendationProbabilities and economic interpretation of cultivar trial results permit growers to distinguish among alternative cultivars with modest differences in expected yield or net return. Use of these methods to interpret test results for growers should improve cultivar selection as well as enhance the benefits of applied research.