Records of 270 meat-type chickens from three lines, 90 of each one, were used to discard variables in a principal component analysis. Data were obtained from meat-type chicken lines of the genetic breeding program of the Universidade Federal de Viçosa. The following traits were evaluated: days at first egg (DFE), egg production rate (EPR) from 22nd to 56th week, body weights at 32nd (BW1), 40th (BW2), 48th (BW3), 56th (BW4), and 64th weeks of age (BW5), and average of three egg weights, at 32nd (EW1), 40th (EW2), 48th (EW3), 56th (EW4) and at 64th weeks (EW5). From the 12 principal components, seven showed variance lower than 0,7 (eigenvalue lower than 0,7), suggesting seven variables to be discarded. Variables which showed the highest coefficients, in absolute value, in the last principal component were discarded. Highly correlated variables with the smaller principal components variance explain a small part of the whole variation. In addition, discarded variables in function of the significant simple linear correlation with the nondiscorded variable, were considered redundants. The variables DFE, EPR, BW4, EW1, and EW4 are recommended for principal component analysis of broiler matrix.
broiler; discard of variable; multivariate analysis; correlation
