About Us | Help Videos | Contact Us | Subscriptions



This article in CS

  1. Vol. 18 No. 5, p. 747-751
    Received: Nov 30, 1977

Request Permissions


Optimal Environments for Yield Testing1

  1. F. L. Allen,
  2. R. E. Comstock and
  3. D. C. Rasmusson2



This research was undertaken to obtain information about the type of yield trial environment that will foster maximum progress from selection. Theory and parameter estimates from five crops that are relevant to the question of the optimum nursery environment for yield testing are reported. The five crop species were barley, Hordeum vulgare L.; wheat, Triticum aestivum L.; oats, Avena sativa L.; soybean, Glycine max (L.) Merr.; and flax, Linum usitatissimum L. Analyses of variance of yield data (a separate analysis for each year and location) reconstructed from information in the annual reports of Uniform Nursery Trials were the sources of our parameter estimates. We defined y as the value of a genotype relative to a test environment, y as the value of a genotype relative to the entire population of environments in which a selected genotype would be used, and H as σ2y/(σ2y + σ2e/n) where σ2y is the variance of y, σ2e, is plot error variance and n is the number of replications in the field trial comparison used as the basis for selection. We then found the proper measure of test environment value to be r√H where r is the correlation between y and y. This is because r√H reflects genotype ✕ environment interaction as well as ✕2e and n, appropriately. The single environment variance analyses available to us provided estimates of ✕2y and ✕2e but not of r. In general, both ✕2y and ✕2e were larger when the nursery yield level was higher. Estimates of √H from the data of low, intermediate and high yielding trials were not greatly different. The indicated advantage (given n = 3) of favorable (high yield) environments over intermediate environments were 4% and 7% for wheat and soybeans, respectively, but in the case of barley, oats and flax, there were no indications that high yield environments were superior. Our results indicate that if any class of environments is eventually established as substantially superior for testing purposes, it will be because the correlation between y and y is comparatively high for that class of environments.

  Please view the pdf by using the Full Text (PDF) link under 'View' to the left.

Copyright © .