Propst, Annabeth L. (1989, ASQC) Process Management Institute, Inc., Bloomington, MN
John Tukey said it best. "Plot the data -- first, last, and always." He was right. Simple plots of the data give us much valuable information that might otherwise be lost. With the increasing availability of computer software, we sometimes forget (or are discouraged by the software) to do these plots.
Not only can simple data plots help us better understand what is happening, but they are easier to explain to the uninitiated. Let's face it -- most of us are not in the enviable position of explaining our results and conclusions to statisticians. We have to convince managers, engineers, first line supervisors, salesmen, and operators. Not only do they not understand the language of statistics, they are often intimidated by it. Showing these people an ANOVA table or telling them that the t-statistic was significant at the .005 level may be less than optimal. How much more convincing could we be if we used simple but effective data plots -- backed up by the statistical tests -- to explain our position.
By doing the statistical tests in conjunction with appropriate plots you can not only learn more about your data, but also explain your results more effectively. The intent of this paper is to illustrate what some of these appropriate plots might be, and what additional information they have to offer. Only a few of the more common statistical techniques will be covered, but the idea applies to all data. Let me plagiarize Tukey and Confucius. "Plot the data -- first, last, and always -- because a picture is worth 10,000 statistics."