Traditional statistical analyses, such as ANOVA and Regression, require that many assumptions be fulfilled in order to obtain accurate results. For example, there are assumptions of normally-distributed data, linear relationships between the dependent variable(s) and independent variables, and homogeneity of variance. In this presentation, we will describe the use of Classification and Regression Trees (CART) to sidestep the assumptions required by traditional analyses. CART has the added benefit of not requiring large sample sizes in order to obtain accurate results, although larger sample sizes are preferred. There is also a difference in the goal of CART compared to traditional analyses. CART is geared toward prediction, whereas traditional analyses are geared toward developing a specific model for your data. The poster will contain specific information about the procedures underlying CART, as well as an example involving data from a legibility of fonts study.

Paper presented to the 4th Annual Symposium on Graduate Research and Scholarly Projects (GRASP) held at the Hughes Metropolitan Complex, Wichita State University, April 25, 2008.Research completed at the Department of Psychology, Fairmount College of Liberal Arts and Sciences