Author
 |
Fisher, Dwight |
|
Submitted to: Journal of Animal Science
Publication Type: Abstract Only Publication Acceptance Date: 7/22/1999 Publication Date: N/A Citation: N/A Interpretive Summary: Technical Abstract: An experimental unit is the unit of experimental material to which a treatment is applied. The definition of the experimental unit has been established since the foundation of agricultural statistics and cannot be logically altered because it forms the basis for analysis. In practice this definition is often overlooked and consequently the experimental unit is sometimes misidentified. The definition of the experimental unit canno be altered without altering the mathematics of the statistical analysis. Agricultural researchers typically use experimental designs and analyses of variance that are predicated upon an accurate identification of the experimental unit. The experimental unit provides the foundation of the definition of experimental error, which is the variation among observations of experimental units treated alike. Randomization is an implied characteristic of distributing treatments among experimental units during the design of a research trial. Lack of randomization can be associated with the misidentification of the experimental unit. For example, in grazing research the group of animals on a pasture makes up a single experimental unit. Each animal on a pasture is a subsample. However, some experimenters conduct an analysis of variance using animals on a pasture as replications. These same researchers would be unlikely to conduct a trial in a barn in which all animals on an experimental treatment were grouped together and yet the two situations are analogous. |
