Penanganan Ketidakhomogenan Ragam Akibat Keberadaan Data Ekstrim Melalui Pendekatan EM-AMMI

Abstract

Multilocation trial is applied to identify adaptation of genotypes from various locations. Additive Main Effect Multiplicative Interaction (AMMI) can identify genotypes that able to adapt in some locations (stable genotype) or genotype that adapt in specific location (specific genotype). Multilocation trial with homogeneity error variance sometimes is not required. So it causes the violation of homogeneity of variance that is part of AMMI analysis. This research uses multilocation trial data from PAIR-BATAN that are consists of 12 genotypes at 5 trial locations. Measurement of responses is dried unhusked rice production (ton ha-1). The exploration of the presence of extreme data shows there are 6 observations of combination genotype by location that have too high variance. The Expectation Maximization-AMMI (EM-AMMI) will solve the violation of homogeneity error variance by eliminating extreme data initially so, AMMI analysis will be used appropriately. AMMI analysis shows that genotype G9 is stable and has a higher average of dried unhusked rice production than both total average and control genotype average.