SPATIAL AUTOCORRELATION IN THE EVALUATIO OF MAIZE COMPOSITES FOR RESISTANCE TO FALL ARMYWORM (Spodoptera frugiperda)

• Published in: Pesquisa agropecuária tropical 2003-07, Vol.33 (2), p.65-72
• Main Authors: dos Santos Reis, AJ, de Miranda Filho, JB
• Format: Article
• Language: Portuguese
• Subjects: Spodoptera frugiperda; Zea mays
• ISSN: 1517-6398

Composites of maize (CRL-01, CRL-02 and CRL-03) were synthesized from crosses of adapted materials with exotic materials that have shown resistance to fall armyworm (Spodoptera frugiperda). Full-sib progenies from the three composites were evaluated for resistance to fall armyworm using a scale varying from 0 (no damage) to 5 (destroyed whorl). The experimental design was randomized complete blocks with two replications. Two models were considered: using the classical analysis with independent errors, and a model with spatially correlated errors. In the spatial model, the residual covariance matrix (R) followed a model built according to the spatial autocorrelation detected in the experiment. The test of Durbin-Watson was used to detect the presence of the spatial autocorrelation between plots, which showed to be highly significant. The distance range of the spatial autocorrelation was about 1.5 m. The use of the spatial model allowed a better local control, resulting a reduction in the estimates of residual variances and an increase of heritability coefficient estimates and expected progress with the selection. The ranking of progenies was changed when using different models of analysis. The use of mixed model was more appropriate than classical analysis in such circumstances.Original Abstract: Foram gerados tres compostos de milho (CRL-01, CRL-02 e CRL-03) entre materiais adaptados e materiais exoticos, com historicos de resistencia a lagarta do cartucho (Spodoptera frugiperda). Familias de iraos germanos dos tres compostos foram avaliadas para resistencia a lagarta, usando uma escala de notas variando de 0 (sem danos) a 5 (cartucho destruido). O delineamento experimental usado foi o de blocos completos casualizados com duas repeticoes. Dois modelos foram considerados: um modelo com erros independentes e outro assumindo erros espacialmente correlacionados (analise espacial). Na analise espacial a matriz de covariancias de residuos (R) foi construida conforme a autocorrelacao espacial detectada em cada experimento. O teste de Durbin-Watson foi usado para verificar presenca da autocorrelacao espacial entre parcelas, a qual foi altamente significativa, O alcance pratico da autocorrelacao espacial foi de cerca 1,5 m. A adocao da analise espacial permitiu uma melhoria no controle da variaco local, resultando numa reducao das estimativas das variancias residuais e, conseqentemente, num aumento dos coeficientes de herdabilidade estimados, com melhorias nos ganh