Statistical significance, selection accuracy, and experimental precision in plant breeding

Genetic selection efficiency is measured by accuracy. Model selection relies on hypothesis testing with effectiveness given by statistical significance (p-value). Estimates of selection accuracy are based on variance parameters and precision. Model selection considers the amount of genetic variability and significance of effects. Questions arise as to which one to use: accuracy or p-value? We show there is a link between the two and both may be used. We derive equations for accuracy in multi-environment trials and determine numbers of repetitions and environments to reach accuracy. We propose a new methodology for accuracy classification based on p-values. This enables a better understanding of the level of accuracy being accepted when certain p-value is used. Accuracy of 90% is associated with p-value of 2%. Use of p-values up to 20% (accuracies above 50%) are acceptable to verify significance of genetic effects. Sample sizes for desired p-values are found via accuracy values.