We found a match
Your institution may have access to this item. Find your institution then sign in to continue.
- Title
New imputation methodologies for genotype-by-environment data: an extensive study of properties of estimators.
- Authors
Angelini, Julia; Cervigni, Gerardo D. L.; Quaglino, Marta B.
- Abstract
The site regression model (SREG) is utilized by plant breeders for analyzing multi-environment trials (MET) to examine the relationships among test environments, genotypes (G), and genotype-by-environment interactions (GE). SREG explores a matrix of G and GE by performing a singular value decomposition on the residuals matrix from a one-way ANOVA, requiring complete data. As missing values are common in MET, we propose two new imputation methods that implement an Expectation Maximization algorithm to fit the SREG model. To evaluate the impact on SREG model parameter estimation of these proposed methods and other competing imputation methods available, we conducted two studies using different approaches. One study involved simulated data while the other used a real dataset. In both studies, different measures to verify whether the joint effect of G plus GE is altered by imputation of data, and the reproducibility of missing data were evaluated. We also incorporated situations not commonly addressed in the literature, such as non-random structures of missing values and a big data situation. The proposed procedures provided estimators with good performance, maintaining superiority in several aspects studied, even when the competing imputation methods did not achieve convergence. Therefore, the new methods enabled incomplete MET data to be effectively analyzed by a SREG model.
- Subjects
MULTIPLE imputation (Statistics); MISSING data (Statistics); SINGULAR value decomposition; GENOTYPE-environment interaction; ONE-way analysis of variance; PLANT breeders; REGRESSION analysis
- Publication
Euphytica, 2024, Vol 220, Issue 6, p1
- ISSN
0014-2336
- Publication type
Article
- DOI
10.1007/s10681-024-03344-z