GRAPHICAL ANALYSIS FOR A LARGE DESIGNED EXPERIMENT

Graphical methods are often used as exploratory tools in analyses of l arge data sets. Their use for exploratory purposes in the context of d esigned experiments is less extensive, however, perhaps because most e xperiments produce relatively small amounts of data. In this article, we show how graphical methods exploiting the structure in a design can be used to reveal interesting patterns in the data and highlight effe cts not necessarily associated with factors in the design. The methods are presented in the context of an analysis of data from a large expe riment (over 3,000 observations) designed to investigate the causes of solder defects on circuit boards.