Use of a partial least squares regression model to predict Test Day of milk, fat and protein yields in dairy goats

A model able to predict missing test day data for milk, fat and protein yields on the basis of few recorded tests was proposed, based on the partial least squares (PLS) regression technique, a multivariate method that is able to solve problems related to high collinearity among predictors. A data set of 1731 lactations of Sarda breed dairy Goats was split into two data sets, one for model estimation and the other for the evaluation of PLS prediction capability. Eight scenarios of simplified recording schemes for fat and protein yields were simulated. Correlations among predicted and observed test day yields were quite high (from 050 to 088 and from 053 to 096 for fat and protein yields, respectively, in the different scenarios). Results highlight great flexibility and accuracy of this multivariate technique.