The goal of this study was to identify a reduced pertinent set of variables from an original data set
of 18 carcass measurements in order to avoid redundancy and collinearity problems or to simplify
data analysis and the development of the linear regression models. Forty-six (46) male lambs, 26
of Churro Galego Bragançano Portuguese local breed and 20 of Suffolk breed were used. Lambs
were slaughtered and carcasses weighed approximately 30 min after in order to obtain hot carcass
weight (HCW). After cooling at 4žC for 24 h a set of seventeen carcass and measurements were
recorded. The data interrelationships common factor analysed following the common factor analysis
procedure. Carcass width and perimeter measurements showed high and positive correlations with
HCW (from 0.74 to 0.91) and between themselves (from 0.55 to 0.80). However, HCW was lowly
correlated with leg length (0.17) and moderately correlated with measurements that characterise
carcass lengths and perimeters (from -0.39 to 0.56). Subcutaneous fat thickness measurements made
at different anatomical positions were lowly correlated with HCW (lower than 0.20), even though
high correlations were observed among the fat thickness measures (higher than 0.67). Four common
factors were retained and identified: carcass weight (factor I), breast bone tissue thickness (factor
II), subcutaneous fat thickness (factor III) and conformation (factor IV), which account for 81.9%
of the variation on the eighteen original variables. This study shows that common factors analysis
can be used to condense the information given by large sets of variables, by selecting a reduced
number of variables, which avoids collinearity problems and simplifies the development of carcass
composition estimation models.
