Performance evaluations of varieties and selections in the Northeastern region of North America

Marketable yield data were extracted from a regional trial system in the Northeastern United States and Canada. Three data bases were formulated containing data from selections (i.e. varieties and breeding lines) evaluated in trials in two consecutive years (1986–87, 1987–88 and 1988–89). Each data base was in the form of an incomplete two-way table as not all selections were tested at all trial locations. The analytic procedure was based on a linear model such that each selection was described by mean marketable yield and a sensitivity measure to environments, and each trial location was described by an environmental effect. An iteration process was used to estimate all parameters. A group of selections was evaluated in all three data bases. Mean marketable yields and sensitivity measures showed fairly good correspondence between data bases. The results provided useful information on selections with regard to their performance and adaptability in the region. A positive association was detected between mean marketable yield and the sensitivity measure, suggesting that a high yielding selection was more sensitive to environmental changes than was a low yielding one. Environmental effects were useful identifying high and low yielding areas. Correlation analyses between observed and fitted yields in individual trials showed that the linear model did not fit well for some trial locations. It is suggested that further research is needed for the investigation of data obtained from a network of trial systems.