Ciência Rural (Aug 2009)
Avaliação de modelos para predição da energia metabolizável do milho para aves Evaluation of models to predict metabolizable energy of corn for chickens
Abstract
O estudo foi realizado com o objetivo de avaliar modelos de predição do valor de energia metabolizável aparente corrigida (EMAn) do milho para aves. Foi organizado um banco de dados contendo informações de experimentos, publicados na literatura nacional entre os anos de 1994 a 2007. Foram utilizados dados de composição química, valores de energia bruta e EMAn de 97 amostras de milho, provenientes de estudos em que os valores de EMAn foram determinados pelo método tradicional de coleta total de excretas com pintos de linhagem de corte. Avaliaram-se cinco modelos, sendo quatro deles sugeridos para estimar a EMAn do milho e um para estimar EMAn de alimentos energéticos. Os modelos estudados foram: EMAnC1 = 36,21*PB + 85,37*EE + 37,26*ENN; EMAnC2 = 37,50*PB + 85,37*EE + 38,21*ENN; EMAnC3 = 4021,8 - 227,5*MM; EMAnC4 = 4337,27 - 57,17*FDN; e EMAnC5 = 4371,18 - 26,48*PB + 30,65*EE - 126,93*MM - 52,26*FB - 25,14*FDN + 24,40*FDA. Os valores de EMAn estimados pelos modelos foram comparados com os valores observados utilizando-se a análise de regressão. Em todos os modelos avaliados, houve rejeição (PA study was realized to validate models to predict apparent metabolizable energy (AMEn) of corn used in chickens diets. Data set with informations of experiments published in national literature among 1994 to 2007 was organized. Chemical composition, gross energy and AMEn of 97 corn samples were including in data set. Studies which AMEn was determined by the traditional method of excreta collection with growing chickens were selected. Five models were evaluates, being four recommended for application in corn and one for application in energetic feedstuffs. The models studied were AMEnC1=36.21*CP + 85.37*EE + 37.26*NNE; AMEnC2 =37.5*CP + 85.37*EE + 38.21*NNE; AMEnC3 =4021.8 - 227.5*ASH; AMEnC4 =4337.27 - 57.17*NDF; e AMEnC5 =4371.18 - 26.48*CP + 30.65*EE - 126.93*ASH - 52.26*CF - 25.14*NDF + 24.40*ADF. The values of AMEn obtained for the models were compared with the observed values using regression analysis. In all the evaluated models there is rejection of the nullity hypothesis, showing differences between estimated and calculated AMEn values. Decomposition of mean squared deviation (MSD) has been demonstrated that calculated values from models AMEnC1, AMEnC3 and AMEnC5 showed that its models failed to simulate the magnitude and pattern of fluctuation. On the other hand, the estimates made with models AMEnC2 e AMEnC4 showed predominance of component bias of prediction that indicates the models failed to simulate the variability of the data around the mean. Energy values predicted using differents models were compared with calculated values by regression analysis. In conclusion any of five models studied allow estimating precisely the apparent metabolizable energy of corn to chickens.
