Abstract
Association between herd sample means and herd sample standard devia- tions for milk yields can be induced by the management practices of culling and selection. This association can be enhanced further by data preparation such as editing short lactations or lactations with low yields. The nature and strength of association in dairy data might be an artifact of editing, selection, and culling rather than evidence that herd yield has a multiplicative effect. This feature of the edited data set must be considered in evaluating analyses, particularly when herd statistics are used to "standardize" production environments.