|Friday, February 22|
|CS09 Theme 4: Software and Graphics #1||
Fri, Feb 22, 1:30 PM - 3:00 PM
Statistics and Graphics for Analysis and InsightView Presentation *David A. Dickey, North Carolina State University
Graphics are useful for many important statistical tasks, among these are exploring data, presenting results, checking assumptions and explaining statistical ideas. The presentation will start with some simple examples of predictive modeling (regression) with graphs used to check residuals, to illustrate the statistical idea of collinearity, and to illustrate interaction. An introduction to seasonal modeling will include a series of graphs showing the effects of the different model components. Turning to generalized linear models, an example of a logistic regression will be presented along with a 3 dimensional plot showing the idea of likelihood maximization and the resulting 95% confidence ellipsoid on the parameter estimates. The presentation will close with some more sophisticated examples, specifically generalized linear mixed models with the emphasis again being on examples and graphics. A Poisson model with random effects will be illustrated with an example on mating habits of horseshoe crabs. An interesting CDC data set on prevalence of active flu virus over several years will be used to show several other models in the generalized linear mixed models category. If time permits, a graph illustrating how logarithmic transformation address multiple problems in data analysis will be presented.