Técnicas estatísticas multivariadas aplicadas a caracterização de carcaça de ovinos da raça morada nova

• Ano de defesa: 2017
• Autor(a) principal: GUEDES, Déborah Galvão Peixôto
• Orientador(a): RIBEIRO, Maria Norma
• Banca de defesa: BRASIL, Lúcia Helena de Albuquerque, MONNERAT, João Paulo Ismério dos Santos, ROCHA, Laura Leandro da, CRUZ, George Rodrigo Beltrão da
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal Rural de Pernambuco
• Programa de Pós-Graduação: Programa de Pós-Graduação em Zootecnia
• Departamento: Departamento de Zootecnia
• País: Brasil
• Palavras-chave em Português: Análise multivariada; Carcaça de ovino; Ovino
• Área do conhecimento CNPq: CIENCIAS AGRARIAS::ZOOTECNIA
• Link de acesso: http://www.tede2.ufrpe.br:8080/tede2/handle/tede2/8127

The aim of this study was to evaluate the application of some of the main techniques of multivariate analysis in a set of variables of carcass traits of Morada Nova sheep breed, in order to reduce the dimensionality of multivariate space, to study the association between group of variables and to select the most important ones and with greater discriminatory power. It was used a data of 48 Morada Nova sheep breed, with a mean age of 8 months, comprising 25 traits regarding carcass measurements (thorax depth - TD, thorax perimeter - TP, leg perimeter - LP, hind perimeter - HP, carcass external length - CEL, carcass internal length - CIL, leg length - LL, hind width - HW, thorax width - TW, index of carcass compactness - ICC, loin eye area - LEA, slaughter body weight - SBW, hot carcass weight - HCW, hot carcass yield - HCY, cold carcass weight - CCW, cold carcass yield - CCY, cooling loss - CL, empty body weight - EBW, true yield - TY, neck yield - NY, shoulder yield - SY, sawcut yield - SCY, loin yield - LY, ribs yield - RY, leg yield - LEY). In the chapter II, 19 variables were submitted to analysis of principal components (TD, TP, LP, HP, CEL, CIL, LL, HW, TW, ICC, LEA, SBW, HCW, HCY, CCW, CCY, CL, EBW, TY) in order to reduce the dimensionality of the data set. The principal components were efficient in reducing the total variation in 19 original variables correlated to five linear combinations (𝐶𝑃𝑘), which explained 80% of the total variation contained in the original variables. The first two principal components together explain 56.12% of the total variance of the variables evaluated. The traits with the highest weighting coefficients, in absolute value, in the first component were CCW (0.37), followed by HCW (0.36), ECW and ICC (0.34), characterizing CP1 as an index for the determination of carcass conformation of the animal. In the second component, the variables LL, HW and HP (0.39) were those with the highest weighting coefficients and indicating that CP2 can be considered an index of the biometric measurements. The variables selected according to the criterion of choice of the one with the highest weighting coefficient in each of the five components were CCW (0.37), HP, LL and HW (0.39), CCY (-0.48), TW (0.50) and LP (0.81), and therefore the use of these traits in future experiments is recommended. In the chapter III, 19 variables (TD, TP, LP, HP, CEL, CIL, LL, HW, TW, ICC, LEA, SBW, HCW, HCY, CCW, CCY, CL, EBW, TY) were submitted to the discriminant analysis to identify the variables with the great discriminatory power between the treatments T1 - forage cactus associated with hay Tifton 85, corn in grains, soybean meal, urea and mineral mix; T2 - forage cactus associated to hay Maniçoba, corn grain, soybean meal, urea and mineral mix; T3 - hay of Tifton 85 grass and 20% concentrate (composed of milled corn, soybean meal and vegetable oil); T4 - hay of Tifton 85 grass and 40% concentrate; T5 - hay of Tifton 85 grass and 60% concentrate; T6 - hay of Tifton 85 grass and 80% concentrate; and also to quantify the association between the variables. Eight variables were selected by the stepwise method: HP, HW, TW, CEL, LL, SBW, EBW and CL. The first three canonical variables were significant, explaining 92.25% of the total variation. The first canonical variable has a canonical correlation coefficient of 0.94, which indicates a high association between the biometric measures and animal performance traits. SBW and HW were the variables selected because they presented the greatest discriminatory power of the treatments, based on standardized canonical coefficients. In the fourth and last chapter, 15 variables were submitted to canonical correlation analysis (TD, TP, LP, HP, CEL, CIL, LL, HW, TW, NY, SY, SCY, LY, RY, LEY) and were shared in two sets (biometric measurements (X) and cut yields (Y)), in order to estimate the canonical correlations between they (𝑊𝑘𝑉𝑘) and evaluate the degree of association between the two groups. Only the first canonical pair was significant with coefficient of 0.86, indicating a high association between the biometric measurements (X) and the meat cut yield traits (Y). The proportion of the shared variance between 𝑊1𝑉1, given by the canonical correlation coefficient squared (r²), was 0.74. That is, 74% of the variation of 𝑊1 is explained by the variation of 𝑉1, which indicates the existence of a high association between the sets of variables X and Y. Considering the standardized canonical coefficients, HW and LL were the variables that have the biggest contribution in the formation of 𝑉1 and SCY and SY were the variables that have the biggest contribution in the formation of 𝑊1. Based on correlation between canonical variable and the original ones for the interpretation of canonical variables, the HW and LL most contributed to perform 𝑉1, whereas SY and SCY were the most important variables to perform 𝑊1.