Abstract:
|
Early in Stat 101, we teach that robustness is important. Yet later in the course, and too often in practice, we ignore those lessons and use simple means and least-squares regression together with Normal-based inferences, even though the corresponding assumptions are violated. Bootstrapping and permutation tests (BPT) let us check the accuracy of common procedures; they are surprisingly inaccurate in the presence of skewness. BPT offer better alternatives, but we need to know what we're doing-the most common bootstrap methods are less accurate than a t-interval for small n. BPT let us more easily do inferences for a wider variety of statistics (e.g., trimmed means, robust regression) and data collected in a variety of ways (e.g., stratification). We'll look at applications from a variety of fields, including telecommunications, finance, and biopharm. BPT provide output we may graph in familiar ways (like histograms) to help students and clients understand sampling variability, standard errors, p-values, and the Central Limit Theorem (CLT)-not just in the abstract, but for the data set and statistic at hand. This course is intended for teachers and practicing statisticians. No familiarity with these methods is assumed.
|