Speaker:
Richard Barfield, PhD
Non-parametric methods provide a convenient way to conduct inference when dealing with a small sample or when necessary data assumptions are not met. Beyond the standard Wilcoxon-style analysis there is a slew of other lesser known non-parametric methods. The coin R package (Conditional Inference, by Hothorn, Hornick, van de Wiel, and Zeiles 2008) provides a set of procedures for various non-parametric tests based on permutation theory. We will discuss the framework motivating the package as well as the procedure used to calculate the exact-permutation p-value. In addition to coin, other unique non-parametric methods will be discussed such as Friedman, Page, and other tests.