Simulation study to assess the performance of EM-AMMI for incomplete G×E data of mango (Mangifera indica) in presence of bienniality

Abstract:

Mango (Mangifera indica) is one of the important perennial fruit crops that exhibits bienniality and results in high economic loss to the growers. In the analysis of Genoype×Environment (G×E) data, it is noticed that few data points are missing.  Due to incompleteness, G×E data becomes unbalanced which is a bit challenging task for the analysis of Genotype ´ Environment Interaction (GEI). The problem gate intensify when crop exihibits bienniality. Expectation Maximization-Additive Main Effect and Multiplicative Interaction (EM-AMMI) is used for imputing missing observations in incomplete G×E data. EM-AMMI involves iterative procedure for filling the missing values in such a manner that starting values do not affect the solution. In present study performance of EM-AMMI method for imputing missing values in incomplete G´E data has been assessed considering four different rates i.e. 5%, 10% 15% and 20% of missing observations in presence of bienniality using original data of mango for four centres as well as simulated data. Computer programme have been developed for generating 1000 data sets for simulation. Number of principal components for imputing missing data using AMMI model has been selected on the basis of Root Mean Square Predictive Distance (RMSPD). Lowest RMSPD with low dispersion was found for AMMI0 at Rewa as well Sangareaddy centre. Hence, AMMI0 was selected as parsimonious AMMI model. Box-plots have also been generated showing the distribution of 1000 correlation coefficients between ranks of genotypes obtained by yield, different stability measures and simultaneously selection of genotypes for yield and stability (SISGYS) indices. It has been found that EM-AMMI does not converge for higher rate of missing observations. Thus, EM-AMMI is recommended for imputation up to 10% of missing observations of G×E data in mango in presence of bienniality. The results have been confirmed by simulation study.