BASUG Quarterly Meeting Announcement
Learn how to get more from your data! Please join us for an informative meeting on data modeling and exploration.
We are delighted to host Daniel Bauer, Associate Professor of Psychology in the LL Thurstone Psychometric Laboratory at the University of North Carolina at Chapel Hill, Dominique Haughton, Professor of Mathematical Sciences at Bentley College, Maria Skaletsky from the Academic Teaching Center at Bentley College, and Timothy C. D’Auria from Creative Computing, Inc.
Topics: Multilevel Modeling in SAS
SAS Text Miner: Introduction and Case study
When: Wednesday, September 24, 2008, 8:45 AM – Noon
Where: Holiday Inn Brookline
1200 Beacon Street, Brookline, MA 02446, 617-277-1200
For directions, please visit: http://www.basug.org/directions.html
How: Individual, On-Line Registration Required. No e-mail please!
To register: Please visit: http://www.basug.org/register.php3
Contact: If you have questions about the meeting contact:
Karen Olson (Karen.Olson@childrens.harvard.edu) or
Eric Hamann (ehamann@wyeth.com)
|
Agenda |
|
|
8:15 - 8:45 |
Sign-in and refreshments |
|
8:45 - 9:00 |
Announcements |
|
9:00 - 10:15 |
Multilevel Modeling in SAS, Daniel Bauer, University of North Carolina at Chapel Hill |
|
10:15 - 10:35 |
Break |
|
10:35 - 11:30 |
SAS Text Miner: Introduction and Case study, Dominique Haughton and Maria Skaletsky, Bentley College |
|
11:30 - 12:00 |
Using SAS Text Mining to Save Lives, Timothy D’Auria, Creative Computing, Inc. |
Note: Times are approximate. Please re-visit the BASUG website (www.basug.org) for updated information.
Abstracts and Bios
Multilevel Modeling in SAS
Daniel Bauer, PhD, University of North Carolina at Chapel Hill
Most statistical models assume independent observations, yet this assumption is often not met in practice. Dependence commonly arises when data are clustered or collected repeatedly over time. For instance, for a given complaint, patients who see the same physician will likely receive more similar treatment than patients who see different physicians, resulting in patient data that is positively correlated within physicians. Such dependence compromises the statistical tests provided by classical modeling approaches, often leading to spuriously significant effects. In contrast, multilevel models (a subclass of mixed models) are specifically designed to model these types of dependent data structures, providing accurate statistical tests and additional information about the outcome under study. This talk will introduce both the key concepts of multilevel modeling and the SAS procedures that can be used to fit linear (MIXED), generalized linear (GLIMMIX), and nonlinear (NLMIXED) multilevel models.
Daniel Bauer, PhD, is an Associate Professor of Psychology in the LL Thurstone Psychometric Laboratory at the University of North Carolina at Chapel Hill. His research focuses on the quantitative modeling of adolescent behavioral health data, including longitudinal data on alcohol and substance abuse, antisocial and criminal behavior, and risky sex. His areas of specialization include multilevel models, structural equation models, and mixture models. In addition to teaching classes in these topics at the university, he developed and delivered a 3-day workshop titled “Multilevel Modeling of Hierarchical and Longitudinal Data Using SAS” through the SAS Education Business Knowledge Series from 2006 to 2008.
SAS Text Miner: Introduction and Case Study
Dominique Haughton, Bentley College and Toulouse School of Economics, and Maria Skaletsky, Bentley College
An Introduction to SAS Text Miner is presented in the context of an analysis of free-text responses to a questionnaire administered to 416 undergraduates in a business university about their perceptions of unfair grading practices. This talk will demonstrate how to cluster responses with SAS Text Miner, and in the process, identify preliminary themes within the responses. Different features of SAS Text Miner will be described as we go along.
A SAS Text Miner Case Study is presented. This study involves a text analysis performed in the context of the database marketing group of a leading provider of information infrastructure. At several contact points, customers provided their job title, but have not always filled in the field “Key Player Description”, which consists of a brief (a few words) job description. The problem here is to impute the missing brief job description for a file of about a million such customers on the basis of available job titles. This talk will demonstrate how this problem was addressed with SAS Text Miner, using a sample of 10,000 customers.
Dominique Haughton, PhD, is a Professor of Mathematical Sciences at Bentley in Waltham, Massachusetts, near Boston. Major areas of interest are applied statistics, statistics and marketing, the analysis of living standards surveys, data mining, and model selection. She is currently researching multilevel models and living standards in Vietnam, clustering methods, models of gambling expenditures and social networks, notably in Senegal. Dr. Haughton is the Editor-in-chief of Case Studies in Business, Industry and Government Statistics (CSBIGS) and Co-editor of The Vietnamese Household: Explorations Using the Living Standards Measurement Survey (1992-1993) and Health and Wealth in Vietnam: An Analysis of Household Living Standards (1998).
Over forty of her articles have appeared in journals such as The American Statistician, BizEd, Telecommunications Policy, Journal of Higher Education Management, Computational Statistics and Data Analysis, Journal of Interactive Marketing, Economic Development and Cultural Change, Studies in Family Planning, Journal of Population Economics, Journal of Biosocial Science, Annals of Statistics, Sankhya, Journal of Statistical Computation and Simulation, Communications in Statistics, Statistica Sinica. She has four current PhD students in Business Analytics.
Maria Skaletsky, MBA, has focused her research on data mining, in particular, web usage and web content mining. Since 2003, she has held the position of research consultant at Bentley’s Academic Technology Center. She is responsible for consulting faculty in all aspects of research approach, design, and implementation, in the selection of appropriate methodology for data gathering and statistical analysis. She also assists faculty in the use of large research databases such as CRSP, Compustat, and Execu Comp, as well as the use of various quantitative and qualitative software such as SAS, SPSS, NVivo and other. Before joining the ATC, Maria received her MBA with a concentration in Business Data Analysis from Bentley College.
Using SAS Text Mining to Save Lives
Timothy C. D’Auria, Creative Computing, Inc.
What if there was a way for a computer to read clinician notes and predict patients most at risk for serious, life-threatening complications before they happen? Unstructured textual data often comprises the vast majority, upwards of 80% in cases, of information available for organizational analysis, particularly in the healthcare industry. During this presentation, we will explore how unstructured data can be incorporated into analyses alongside structured sources to predict future patient emergency room usage. We will then discuss why text mining may be a preferable alternative to structured healthcare data analysis using logistic regression and other statistical procedures. The presentation will conclude with a summary of how SAS text mining capabilities can ultimately be leveraged to save patient lives.
Timothy C. D’Auria leads the SAS Analytics Practice at Creative Computing, Inc., a SAS Silver Consulting Partner. As a patented and published author with distinction in predictive modeling technologies, Timothy’s interest is in the application of data mining to improve patient care and achieve competitive advantage in healthcare industries. His work in business data mining has been featured in leading industry publications and has been incorporated into educational programs at NYU, the University of Denver, and Cornell University. He also led the successful development of three analysis-driven businesses from the ground-up, one of which has become a leading provider of hospitality forecasting technologies. His 3-Day Analytical Discovery Workshop helps organizations identify and develop opportunities to strategically leverage statistics and data mining to create change and improve service. Timothy received his Bachelors of Science degree in Statistics and Biology with Distinction in Research from Cornell University and is a certified Emergency Medical Technician. His most recent article, “What are Analytics?” is currently in-press with Managed Healthcare Executive.
BASUG Membership
Keep your BASUG Membership up-to-date! Bring a completed membership form to the meeting, along with your check. The Individual Membership fee is $30 per year. Corporate memberships were discontinued. For more information on our membership policy, or to print a membership form visit: http://www.basug.org/basugj.shtml
Directions and Parking http://www.basug.org/directions.html
BASUG Contacts To email our Webmaster: basugwm@basug.org
Mailing Address: BASUG
PO Box 253
Boston, MA 02117