Premium
Simultaneous Selection for Yield and Stability in Crop Performance Trials: Consequences for Growers
Author(s) -
Kang Manjit S.
Publication year - 1993
Publication title -
agronomy journal
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.752
H-Index - 131
eISSN - 1435-0645
pISSN - 0002-1962
DOI - 10.2134/agronj1993.00021962008500030042x
Subject(s) - stability (learning theory) , type i and type ii errors , statistics , mathematics , yield (engineering) , statistic , selection (genetic algorithm) , crop , pairwise comparison , interaction , variance (accounting) , agronomy , biology , computer science , economics , machine learning , materials science , artificial intelligence , metallurgy , accounting
Utilization of genotype × environment (GE) interaction encountered in crop performance trials is an important issue among plant breeders and agronomists. Practical integration of yield and stability of performance has not been achieved. The purposes of this paper are (i) to examine consequences to growers when researchers commit Type I (rejecting the null hypothesis or H o when it is true) and Type II (accepting the H o when it is false) errors under a yield‐based, conventional selection method (CM) and a proposed method (yield‐stability statistic or YS i ) that uses GE interaction, and (ii) to show why a greater emphasis on the stability component would be advantageous to growers. A corn ( Zea mays L.) dataset was used to compute the YS i and to estimate Type II error rates for overall mean yield comparisons and for the stability‐variance statistic, σ 2 i (σ 2 i measures contribution of i th genotype to the total GE interaction) at different α (Type I error rate) and δ (minimal detectable difference) levels. When pairwise yield comparisons are made, a higher level of a will not be as harmful to growers as a higher level of β (Type II error rate). Since researchers would prefer to have power (1 ‐ β) of a test between 0.70 and 0.80, choosing an α(1) level (one‐tailed) between 0.10 and 0.20 should be appropriate. If an α(1) value of 0.25 is selected δ between 1.0 and 1.4, the Type II error rate would be almost zero. For stability, the H o tested was that yield means of a genotype in different environments were equal, i.e., σ 2 i = 0. A consequence of committing a Type I error would be that growers could miss using a stable cultivar, but a consequence of committing a Type II error can be disastrous for growers, as they could choose an unstable genotype and suffer economically. A greater emphasis on performance stability during selection would benefit growers.