Scientific Program > Topic 7 > Session 7G >
 Presentation 7G3. Using Fathom to Promote Dynamic Explorations of Statistical Concepts

 Presenter Robin Lock (USA) rlock@stlawu.edu

 Presentation Abstract Although the use of a statistical computer package has become an integral part of many modern statistics courses, the primary goal of traditional software has been to do statistics rather than to learn statistics. Fathom: Dynamic Statistics Software is one of several newly developed packages that focus a greater emphasis on providing an atmosphere in which students can explore statistical concepts and ideas. To facilitate learning, a key premise of the software is that all aspects of an analysis are linked so that students can see how changes in one area are reflected in another. For example, one might look at a scatterplot with a regression line, correlation, and test for linear association and see how all of those items behave when an influential point is dragged around the plot. Fathom's developers have made a special effort to produce an intuitive interface that allows students to "drag & drop" to construct analyses from basic building blocks. One of Fathom's unique features is the ease with which a student can collect the values for a statistical quantity across many simulated samples. While this provides yet another tool for demonstrating the central limit theorem for a sample mean, one can just as easily look at the sampling distribution of the median or create one's own statistic. What if we average the mean, median, and midrange to estimate the center of a distribution? Although Fathom is an easy package for students to learn to use on their own, it also provides a convenient environment for instructors to develop their own demonstrations. For example, one can attach the parameters of a density curve such as the normal to Fathom "sliders" that allow students to change parameters dynamically to investigate how they relate to the shape of the curve. A histogram of sample proportions could be linked to sliders for the sample size and population proportion. One might also create a scatterplot with a given correlation as a slider and then have the plot updated as the correlation moves from +1 to -1 in a smooth animation. Techniques needed to develop these sorts of dynamic illustrations will be discussed and demonstrated.