Stability Analysis for a Countrywide Series of Wheat Trials in Pakistan. Issue 5 (1st September 2016)

• Record Type: Journal Article
• Authors: Piepho, Hans‐Peter; Nazir, Mian Faisal; Qamar, Maqsood; Rattu, Atiq‐ur‐Rehman; Riaz‐ud‐Din, ; Hussain, Manzoor; Ahmad, Gulzar; Fazal‐e‐Subhan, ; Ahmad, Javed; Abdullah, ; Laghari, Karim Bux; Vistro, Imad Ali; Kakar, M. Sharif; Sial, Mehboob Ali; Imtiaz, Muhammad
• Is Part Of: Crop science. Volume 56:Issue 5(2016)
• Page Start: 2465
• Page End: 2475
• Publication Date: 2016-09-01
• Subjects: Crop science -- Periodicals; Cultures -- Périodiques; Cultures de plein champ -- Périodiques; Crop science; Nutzpflanzen; Zeitschrift; Pflanzenbau; Periodicals; 633
• DOI: 10.2135/cropsci2015.12.0743
• Languages: English
• ISSNs: 0011-183X

Abstract

Abstract : In Pakistan, wheat ( Triticum aestivum L.) varieties undergo intensive yield testing within the National Uniform Wheat Yield Trials (NUWYT) before recommendations are made to farmers. The trialing network extends across 12 different agroecological zones defined for the whole of the country. In this paper, we consider the analysis of data from this trial system over 2 yr and 34 locations. For each of the 16 varieties, variances of genotype × location (G × L) and genotype × location × year (G × L × Y) interactions were estimated to assess stability. Substantial differences in stability were found between the 16 varieties. The fitted model was used further to compute genotype means for each of the seven irrigated agroecological zones. It is demonstrated that a model with random genotype × zone (G × Z) effects can be used to borrow strength across zones, which is particularly beneficial for zones with a small number of test locations. Our results show considerable shrinkage of best linear unbiased predictions (BLUPs) of G × Z means compared with best linear unbiased estimators (BLUEs), which reflects a considerable amount of environmental noise in the data. A salient feature of our approach is that shrinkage depends on stability variances, thus providing a convenient insurance mechanism against overoptimistic as well as overpessimistic mean yield estimates for particularly unstable varieties. At the same time, this approach provides a new way to combine stability and mean performance in a meaningful way.