Modelos de predição para sobrevivência de plantas de Eucalyptus grandis / Prediction models of Eucalyptus grandis plant survival

Objetivou-se com este trabalho comparar modelos de predição de plantas sobreviventes de Eucalyptus grandis. Utilizaram-se os seguintes modelos: modelo linear misto com os dados transformados, utilizando-se as transformações angular e BOX-COX; modelo linear generalizado misto com distribuição binomial e funções de ligação logística, probit e complemento log-log; modelo linear generalizado misto com distribuição Poisson e função de ligação logarítmica. Os dados são provenientes de um experimento em blocos ao acaso, para avaliação de progênies maternas de Eucalyptus grandis, aos 5 anos de idade, em que a variável resposta são plantas sobreviventes. Para comparação dos efeitos entre os modelos foram estimadas as correlações de Spearman e aplicado o teste de permutação de Fisher. Foi possível concluir que, o modelo linear generalizado misto com distribuição Poisson e função de ligação logarítmica se ajustou mal aos dados e que as estimativas para os efeitos fixos e predição para os efeitos aleatórios, não se diferenciaram entre os demais modelos estudados.

The objective of this work was to compare models for prediction of the survival of plants of Eucalyptus grandis. The following models were used: linear mixed model with the transformed data, by utilizing the angular transformations and BOX-COX; generalized linear mixed model with binomial distribution and logistic functions, probit and complement log-log links; generalized linear mixed model with Poisson distribution and logarithmic link function. The data came from a randomized block experiment for evaluation of Eucalyptus grandis maternal progenies at five years old, in which the variable response are surviving plants. For comparison of the effects among the models the correlations of Spearman were estimated and the test of permutation of Fisher was applied. It was possible to conclude that: the generalized linear mixed model with Poisson distribution and logarithmic link function misadjusted to the data; the estimates for the fixed effects and prediction for the random effects did not differ among the to other studied models.

Distribution of aflatoxins in corn fractions visually segregated for defects

The aflatoxin distribution in corn fractions obtained after visual segregation for defects in 30 samples, known to be contaminated, was studied. Each sample was passed through a 5.0 mm round holes sieve, graded for defects and then segregated in sound kernels (regular kernels) and non-sound kernels (injured, germinated, fermented, moldy, heated, insect damaged, immature, broken, hollow, fermented up to », discolored, extraneous materials, and injured by other causes), as defined by the Brazilian Official Grading rules for corn. The non-sound kernels showed the highest contamination levels in all samples. The contamination levels of non-sound kernels (20 percent of total weight) ranged from 23 to 1,365 æg/kg of aflatoxins (B1, B2, G1 and G2) and were higher than sound kernels (p<1 percent) ranging from not detected (ND) to 126 æg/kg and in 87 percent of these the aflatoxin contents were lower than 20 æg/kg. Statistically significant correlation indexes were found among the percentage of defective groups like fermented, heated and sprouted kernels or the total injured kernels, and the estimated contamination levels for the sound and non sound fractions. It was concluded that the non-sound kernels fraction, even being small in weight, has contributed with 84 percent of the estimated contamination of the samples. The segregation of the non-sound kernels would favor a reduction in the contamination of corn lots. The poorer quality corn types (types 3 and Bellow Standart) have predominated among samples of the experiment

Disribution of aflatoxins in comtaminated corn fractions segregated by size

The efficiency of segregation, by particle size, in reducing the aflatoxin content of corn lots under the Brazilian conditions was studied. The aflatoxin content, in the corn fractions obtained after sieving thirty contaminated samples, with approximately 3.0 kg each, was determined. The samples were passed through 4.5 mm round-hole sieves and the fractions with particles ü4.5 mm (fraction A) and <4.5 mm (fraction B) and then analyzed for aflatoxins by TLC. The results showed that the distribution of the aflatoxins among the fractions A and B varied from sample to sample. However, the fraction B presented higher content (6 to 1,422 µg/kg) of aflatoxins (P<5 per cent), than samples of fraction A (4 to 389 µg/kg). In spite of that the segregation of this fraction did not reduce significantly the contamination of the samples. This probably occurred because the contribution of the fraction B for the estimated total of the concentration of aflatoxin of the sample was smaller than that of the fraction A, due to the smallest amount of the fraction B in the sample weight total. In spite of being significant for the studied samples the correlation among the percentages of the groups of grains with defects a) burning grains, sprouted, burned and moldy and b) total injured, with the estimated level of contamination for the samples, more studies will be necessary so that better conclusions can be drawn for this correlation.