Abstract
The accuracy and precision of various IR techniques for use with mild components was studied. Milk fat was determined at wavelengths 3.48 and 5.73 .mu.m and by a combination of these 2. Fat and protein determinations were compared both with and without simultaneous lactose determintion. The interference of secondary components and their influence on accuracy were also studied. The standard deviation (accuracy) of milk fat analyses at wavelength 3.48 .mu.m was similar to that at 5.73 .mu.m, .apprx. 0.08 percentage units, but the combination of these wavelengths gave the lowest standard deviation (i.e., greatest accuracy). Determination at wavelength 3.48 .mu.m was not particularly dependent on variations in refractive index, but analyses at this wavelength had instead a lower repeatability. When using a constant of lactose secondary interference in fat and protein analyses, it was the protein determinations that were mostly affected. Secondary interference coefficients showed breed dependence and had quite a profound effect on accuracy, especially in milk having a widely varying milk components as in Jersey milk. Theorectical derivations for the influence on accuracy of secondary interference constants was provided.