Klasifikasi Genotipe Pada Data Tidak Lengkap Dengan Pendekatan Model Ammi

Abstract

Multilocations trials play an important role in plant breeding and agronomic research. Study concerning genotype-environment interaction needed in selection of genotype to be released. AMMI (Additive Main effect and Multiplicative Interaction) is one of statistical technique to analyze data from multilocations trials. The analysis of AMMI is a combining analysis between additive main effect and principal component analysis. The main restriction of using AMMI analysis is balance data. However a multilocations trials give an opportunity of the occurrence of unbalance data become very big. It is intended for every combination of genotype and location have the same number of replication. Therefore, we must estimate the data which do not complete. Incomplete data case, it is needed some data estimation method analysis, at this research employed connected data method and EM-AMMI algorithm to estimate incomplete data. Data which used in this research is obtained from Indon Spices Medicinal Crops Research Institute (ISMECRI) Bogor. It is a secondary data of ginger essential content resulted from multilocations trials, sixteen of genotype ginger which tested at five location in West Java. Simulation result shows that estimation using connected data will obtain smaller MAPE if its followed by two times EM-AMMI algorithm, for the percentage of data lose 10% to 20%. Production of ginger essential content estimated with AMMI2 model can explain 76.09% interaction structure among location and genotype. Seen from stability of genotype, there are 2 stable genotype, that is genotype 3 and 7. Genotype 8, 12, 15 specific at location B; genotype 9, 11, 14, 16 specific at location D and E; and genotype 1, 2, 4, 5, 6, 10 and 13 specific at location A and C.