THE UTILIZATION OF MULTIDIMENSIONAL-SCALING TO IDENTIFY FORAGE CHARACTERISTICS ASSOCIATED WITH PREFERENCE IN SHEEP

Forage preference is difficult to assess but is likely an important fa ctor in determining DM intake. In two experiments, six sheep (wethers, Exp. 1; ewes, Exp. 2) were used to obtain preference ratings on nine hays (Exp. 1) or nine fresh forages (Exp. 2). The statistical procedur e of multidimensional scaling was used to develop orthogonal dimension s to account for the observed variation in preference for hays and fre sh forages. This analysis produced three dimensions that accounted for 54% of the variation in mean 3-d preference for hays (Exp. 1). The th ree dimensions were subsequently associated by regression analysis wit h seven compositional variables. These were IVDMD and concentrations o f monosaccharides, short-chain polysaccharides (SCP), and disaccharide s in dimension 1 (R-2 = .99); acid detergent insoluble ash (ADIA) conc entration, monosaccharide and SCP concentrations, and masticate NDF in dimension 2 (R-2 = .99); and disaccharide and starch concentrations i n dimension 3 (R-2 = .65). Three dimensions accounting for 51% of the variation in fresh forage preference were found and related to four va riables by regression analysis. These were disaccharide concentration and median particle size in dimension 1 (R-2 = .76) and starch and mon osaccharide + disaccharide concentrations in dimension 3 (R-2 = .97). No physicochemical variables analyzed in this study were found to be s ignificantly associated with the coordinates for the forages in dimens ion 2. Multidimensional scaling accounted for more than half the obser ved variation in animal preference ana, in all but one case, these dim ensions could be associated with physicochemical characteristics of th e forages.