Students first learn statistical thinking through simple graphics like bar charts and histograms. At the college level, a graphical approach to statistics is less often used. This is so in part because good graphics have only just become available on computers. Possibly a more important reason is that while numerical methods have a firm mathematical footing, graphical statistics have consisted primarily of methods of graphing, without accompanying theory. In this research, the project is developing a theoretically sound graphical approach to regression analysis, which is the collection of statistical techniques designed to understand how one variable depends on others. What differentiates this project from others on graphics is the development of a unified theory to make graphs more comprehensible, and much more useful in the analysis of data. The project is producing self-contained teaching materials, textbooks and teacher's aids, and presenting workshops for college teachers on how to use this material.