All observations (data) contain errors. Understanding the sources of these errors is important to reach the correct decision from the data, or else you risk driving into smoke. Some sources of errors are linked to the physical limitations of the measuring devices. This is the type of errors that people working in the physical sciences are accustomed to. Reporting data with more digits than what is legitimate from the precision of the instrument is frequent, but very misleading. People working with live things, such as cows, must understand that data also contain errors because living entities vary. For example, the milk production and body weight of a given cow continuously vary. The sizes of the daily variation of many traits within a cow are suchthat little can be inferred from one single datum. In addition, there is variation amongst animals treated alike, which is the basis of replication in research. Because cows within a pen are not independent, any factors common to a pen will affect all animals within it. Looking at feed analyses, data contain errors (variation) that are intrinsic to the feed (i.e., true), and errors that are due to the observer. In most instances, the sampling variation in forages is such that little can be inferred froma single sample. Much progress would be made if 2 independent samples were taken and assayed each time a nutritionist need data on feed composition.
展开▼
