Sources of error

In any experiment, there will be different sources of variation, often also more commonly called error. A key to understanding the analysis of designed experiments using regression is to be able to calculate the magnitude of each source of error or variation. The majority of textbooks and statistical packages use multilinear models and are based on principles first established around a century ago by pioneers of experimental design such as Fisher. Box and co‐workers developed such approaches in the 1950s to 1970s primarily for a statistical audience, and Deming and Morgan carried the torch primarily for analytical chemists in the 1970s to 1990s. There are several books that go into the mathematical details published over the last few decades. Bayne and Rubin's text is recommended for further reading among others.