Energia metabolizável de alimentos energéticos para suínos: predição via meta-análise, determinação e validação por simulação bootstrap
| Ano de defesa: | 2013 |
|---|---|
| Autor(a) principal: |
Langer, Carolina Natali
 |
| Orientador(a): |
Oliveira, Newton Tavares Escocard de
|
| Banca de defesa: |
Nunes, Ricardo Vianna
,
Rossi, Robson Marcelo
|
| Tipo de documento: | Dissertação |
| Tipo de acesso: | Acesso aberto |
| Idioma: | por |
| Instituição de defesa: |
Universidade Estadual do Oeste do Paraná
Marechal Cândido Rondon |
| Programa de Pós-Graduação: |
Programa de Pós-Graduação em Zootecnia
|
| Departamento: |
Centro de Ciências Agrárias
|
| País: |
BR
|
| Palavras-chave em Português: | |
| Palavras-chave em Inglês: | |
| Área do conhecimento CNPq: | |
| Link de acesso: | http://tede.unioeste.br:8080/tede/handle/tede/1538 |
Resumo: | The proposed objectives in this study were the metabolizable energy (ME) prediction of corn, sorghum and wheat bran from the chemical and energy composition of these foods in national and international literature data; the stepwise procedure validation of regressive selection by bootstrap simulation; the ME determination of these foods for growing pigs and subsequent validation of equations estimated in ME values observed in the experiment, using the bootstrap resampling procedure. For the ME prediction in chemical composition function, we used data from trials of pig metabolism and chemical composition of corn, sorghum and wheat bran, available in national and international scientific literature. Five models of multiple linear regression were adjusted to estimate the ME. In the stepwise procedure validation of regressive selection, it was used the non-parametric bootstrap resampling method, with each sample replacement, from the database formed via meta-analysis. It was observed the significance percentage by regressive (SPR) and the joint occurrence percentage of the model regressive (JOPMR). In the complete model and in the model without the digestible energy inclusion (DE), the DE and the gross energy (GE) were the regressive which presented the highest SPR (DE = 100% and GE = 95.7%), respectively, suggesting the importance of such regressive to explain the ME of energetic foods for pigs. However, the JOPMR were low, with values among 2.6 and 23.4%, indicating a low reliability of the predicted models to estimate the ME of corn, sorghum and wheat bran for pigs. Based on the SPR, the regressive of the models ME4 = 3824.440 - 105.294Ash + 45.008EE - 37.257DA1*CP (R2 = 0.90); ME5 = 3982.994 - 79.970Ash - 44.778DA1*CP - 43.416DA2*Ash (R2 = 0.92) are valid to estimate the ME of energetic food for pigs. In the field trial, we used 44 crossbred pigs, male and castrated, with an average initial weight of 24.3 ± 1.12 kg, in a randomized block experimental design, with ten treatments and a reference ratio. The ten treatments consisted of six corn and two sorghum cultivars, which replaced in 30% the RR, and two wheat brans, which replaced 20% of the RR. The method of total collection of feces and urine was used for determining the ME of food by using ferric oxide as a fecal marker to define the beginning and end of the collection period. The ME values of corn, sorghum and wheat bran for pigs vary from 3.161 to 3.275, from 3.317 to 3.457 and from 2.767 to 2.842 kcal kg-1 of natural matter, respectively. The validation of the ME prediction models was performed through adjusting the linear regression models of 1st degree from the observed values experimentally determined in function of ME predicted values, calculated by replacement of chemical and energetic composition values of foods, determined in laboratory, in the estimated models via meta-analysis, using the ordinary minimum squares method. The validation of 1st degree models and prediction models of ME was verified by testing the joint null hypothesis for the linear regression parameters (H0: β0 = 0 and β1 = 1). The crossvalidation percentage of each estimated model was evaluated by the same validation tests described in the single test validation. The model ME1 generated similar predicted EM values (p>0.05) to the experimentally observed ME values for national corn and sorghum cultivars in single test validation and had the highest percentage of validation (68%) in 200 bootstrap samples. The other models had a low percentage of cross-validation (0 to 29.5%), and the validated model by both procedures, and that can be used for national corn and sorghum is the ME1 = 2.547 + 0.969DE |