Plant breeding decisions based on different amounts of data: judicious use of replication
In order to allocate fixed resources in a plant breeding program to maximize efficiency, a balance must be achieved between the accuracy of estimation of a trait and the number of estimations conducted. Typically, greater accuracy of estimation is achieved through replicated measurements of a trait or a more refined analysis. However, these options come at the cost of reducing the number of different lines that can be evaluated. Decisions based on different amounts of yield data (number of replicates, locations) were compared. Using three years of data, the comparison of one replicate at two locations to two replicates and two locations was found to retain about 80% of the decisions in common. Under this scenario, twice as many entries can be evaluated using the same plot resources while risking 20% more wrong decisions. A stronger selection intensity that would follow with a larger number of entries may ameliorate some of the 20% questionable decisions. This study of yield estimations in the U of S durum breeding program is used to modify the structure of the program to maximize the likelihood of identifying a superior cultivar.
The following license files are associated with this item: