ANNOTATIONS TO THE STATISTICAL EVALUATION OF SPLIT-SAMPLE TRIALS ON THE INFLUENCE OF DIFFERENT TREATMENTS ON SEMEN QUALITY

Results from split-sample-trials based on relatively high numbers of e jaculates from a limited number of animals are often required to be va lid for the whole population from which the experimental animals origi nate. As a prerequisite, certain rules concerning the statistical eval uation of data have to be considered. It is stated that statistical co mputation based on the means over the ejaculates for any animal leads to fewer errors and is indicated more often than a computation based o n single ejaculates. Using linear models, it is shown that a lower num ber of ejaculates from a high number of animals is often preferable to higher numbers of ejaculates from a lower number of animals.