Premium
Interrelationships among and Repeatability of Several Stability Statistics Estimated from International Maize Trials
Author(s) -
Pham H. N.,
Kang M. S.
Publication year - 1988
Publication title -
crop science
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.76
H-Index - 147
eISSN - 1435-0653
pISSN - 0011-183X
DOI - 10.2135/cropsci1988.0011183x002800060010x
Subject(s) - statistics , repeatability , statistic , biology , linear regression , summary statistics , regression , zea mays , rank (graph theory) , mathematics , agronomy , combinatorics
Importance and implications of genotype‐by‐environment interaction are well known to breeders and geneticists. Several methods of measuring stability of performance of genotypes across environments have been developed. Some of these methods have been compared empirically and statistically. The purpose of this investigation was to study rank correlations among Eberhart and Russells regression coefficient (b i ‐ER) and deviation from regression (sd 2 1 ); Shuklas σ 2 1 , s 2 1 , and b i ‐Sh; CV i ; and S 2 i statistics. The data used for calculating these statistics were obtained from five international maize ( Zea mays L.) trials. Trial 1 involved 14 genotypes tested at 39 locations in 21 countries, trial 2 involved 23 genotypes tested at 46 locations in 27 countries, trial 3 involved 16 genotypes tested at 23 locations in 13 countries, trial 4 involved 18 genotypes tested at 25 locations in 15 countries, and trial 5 involved 12 genotypes tested at 17 locations in 10 countries. The statistic sd 2 i was highly rank correlated with σ 2 i and s 2 i , but not with other statistics. The b i ‐Sh and b i — ER statistics were highly correlated (rank) with each other as well as with S 2 i . None of the statistics was consistently well correlated with mean yield. Stability statistics were also calculated by grouping locations for a trial into high‐yield (top 50%) and low‐yield (bottom 50%) environments. Repeatability of all statistics between the two groups was generally negligible. Repeatability was also negligible between statistics estimated from each of two randomly chosen subsets of environments and from four randomly chosen subsets of environments. The high‐yield environments were characterized by relatively low CV i and the low‐yield environments by relatively high CV i .