Planning and Analyzing Experiments with Models that Distinguish Between Replicates and Repeats

A commonly used model to analyze experiments with normal responses does not distinguish between replicates and repeats. The same problem arises with binary and count responses where we can use a generalized linear model. In this article, we propose using models that explicitly allow for two sources of variation, that due to replicates and that due to repeats. In addition, for experiments carried out on high‐volume, existing processes, there are often large amounts of data, collected in different ways, that are available to aid in the planning and analysis of the experiment. We demonstrate the value of using these available data with two detailed examples. We finish with a brief summary and raise some further issues.