Fall 2014 SCASA Kick‐off Meeting

Click here to download the schedule of events.
 

• Location: Argyros Auditorum, Arnold and Mabel Beckman Center, City of Hope, 1500 E.  Duarte Road, Duarte California, 91010‐3000  
• Keynote Speaker: Sudipto Banarjee, Professor and Chair, UCLA Department of  Biostatistics  
• Schedule:  

 o 12:30‐1:00:  Coffee, Tea, Cookies and Conversation  
 o   1:00‐2:00:  Student and Industry Speakers
 o   2:00‐3:00:  Keynote Address  
 o   3:00 ‐ end:  Business Meeting  

•  Business Meeting Agenda:  

o Nomination of President Elect, Secretary, VP for Membership, Treasurer  (Catherine Sugar)  
o Treasurer’s Report (Jay Bartroff)  
o Report on the Bob Newcomb Memorial (Andy Dagis)  
o Career Day Reminder—City of Hope, February 14th, 2015 (Andy Dagis)  
o Update on the Applied Statistics Workshop (Olga Korosteleva/Harold Dyck)
o Update from SCASA Membership Committee (Karabi Nandy/Mike Jay)
o Brainstorming for Spring Poster Competition (Heidi Gransar)  
o Open Floor 

• Keynote Talk:  Bayesian Modeling and Inference for Large Geographically Referenced Data  Sets 

• Abstract: With the growing capabilities of Geographical Information Systems (GIS) and user‐friendly  software, statisticians and data analysts today routinely encounter geographically referenced datasets  containing observations from a large number of spatial locations. Over the last decade, hierarchical spatial  process models have become widely deployed statistical tools for researchers to better understanding the  complex nature of spatial variability and developing predictive frameworks. It is little surprise that such  models have found a wide range of applications from interpolating levels of pollution in the atmosphere to  predicting real estate home prices. However, fitting hierarchical spatial models is an expensive exercise and  becomes impractical when the number of spatial locations is large. In this talk, I will present a couple of  approaches for constructing rich and flexible spatial models especially geared toward analyzing large datasets.  These approaches use concepts such as dimension‐reduction and sparsity to achieve substantial  computational savings. The algorithms used to estimate the statistical models are based on popular Markov  chain Monte Carlo simulation methods but are constructed carefully to save on storage and computer  operations. We compare and assess the relative merits and demerits of these methods and demonstrate their use in inference on the spatial distribution of forest biomass from a massive US Forest Inventory database  spanning the continental US. 
  
  

• About the Speaker: Sudipto Banarjee, Ph.D., is Professor and Chair of the UCLA Department of  Biostatistics where he moved this past summer after spending 14 years on the faculty of the  University of Minnesota. He received an M.Stat degree from the Indian Statistical Institute and  his M.S. and Ph.D. from the University of Connecticut, Stores. Dr. Banarjee’s research focuses  on statistical modeling and analysis of geographically referenced datasets, Bayesian statistics,  the interface between statistics and Geographical Information Systems, and statistical  computing. He has published over eighty peer‐reviewed journal articles and has co‐authored a  book entitled "Hierarchical Modeling and Analysis for Spatial Data". A fellow of both the  American Statistical Association and the International Statistical Institute, Dr. Banarjee has  been honored with the Abdel El Sharaawi Award from the International Environmetrics Society  (2009) and the Mortimer Spiegelman Award from the American Association of Public Health  (2011). This latter prize, given annually since 1970, is presented to the nation's most  outstanding public health statistician under the age of 40. Dr. Banarjee sits on numerous  journal editorial boards and grant review panels, including serving as a permanent member of  the NIH Biostatistical Methods Research and Design Study Section.