This is the program for the 2010 Joint Statistical Meetings in Vancouver, British Columbia.
|CE_35T||Wed, 8/4/2010, 3:00 PM - 4:45 PM||CC-11(East)|
|Statistical Sensitivity and Graphical Methods for Correlated Data Analysis — Continuing Education CTW|
|ASA , Data Numerica Institute|
|Instructor(s): Edward C. Chao, Data Numerica Institute, Inc.|
|This course aims at the analytic, sensitivity and graphical methods for correlated data, especially longitudinal data with possibly missing outcomes. The response could be continuous or discrete, and the data could be complete or incomplete. We will discuss a set of modeling approaches, including Linear Mixed-effects Models (LME), Generalized Linear Mixed Models (GLMM), and Generalized Estimating Equations (GEE) Models. To study correlated data, it is important to apply appropriate methods, and we will demonstrate with real cases. To explore data and to investigate the validity of analytic results, graphical methods and sensitivity analyses are quite important. These methods provide data analysts a set of tools to investigate the impacts and potential departures from the modeling methods and assumptions. One challenging area occurs when the mechanism of missing data depends on the unobserved outcomes, i.e. non-ignorable missing. In longitudinal studies, missing data could occur when subjects leave the study due to outcome-related or unrelated causes such as side effects in medical studies. Software in R and web application with friendly user interface and dynamic graphics will be presented. We will also compare software and provide suggestions. Some knowledge in the analysis of correlated data is useful but not required.|