Organizing Committee

Russell Steele

(McGill University)

Geneviève Lefebvre

(Université du Québec à Montréal)

CRM Statistics Laboratory

 

Contact

Guillermo Martinez-Zalce

CRM Laboratories Coordinator
514-343-7574

October 22, 2010

Missing Data Approaches in the Health and Social Sciences: A Modern Survey

Much of modern research in medicine and the social sciences requires the analysis of large databases of information. Although such databases are extremely valuable due to the immense quantity of information, they present challenges to data analysts when subjects' records contain incomplete (or missing) information. There is a great divide between the statistics research community and the non-statistical research community with respect to the methods used to analyze such data.  Although statisticians have devised methods with statistically correct and efficient properties, many of these approaches have been unable to gain traction in the general community. Statisticians have proposed two different types of approaches:  imputation of the missing values and weighting methods.   Both of these approaches have relative advantages and disadvantages, but these have not often been jointly addressed in published research.  The disagreements in the statistics community have led to confusion in the substantive community and even applied statisticians are unclear as to what constitutes correct, but practical approaches to complicated missing data problems.
 
This workshop will contain five talks that broadly cover the most popular approaches to analysis of missing data in the medical and social sciences. Grace Yi (University of Waterloo) is the winner of the 2010 CRM-SSC prize and has contributed in a significant way to the development of statistical methods for longitudinal studies and for the analysis of time-to-event data, especially for the treatment of missing observations and measurement error.  James Carpenter (London School of Hygiene and Tropical Medicine) has done extremely important work, not only in statistics, but also in the area of public health in the United Kingdom, demystifying advanced statistical methods for missing data.  In particular, he was co-author of a report to the National Institute for Health Research in the United Kingdom on the analysis of missing data in randomized clinical trials.  Ofer Harel (University of Connecticut) is an expert in multiple imputation methods for missing data and measurement error, with an emphasis on applications in medicine. Two local researchers will also present talks at the workshop. David Haziza (Université de Montréal) and Michael Regier (McGill) work in the areas of missing data in survey analysis and maximum likelihood methods for missing observations, respectively.

Schedule in brief

Venue: UQAM, Pavillon Sherbrooke,
200 rue Sherbrooke O.,
Montréal, Qc H2X 3P2

Conferences: Room SH-3260
Reception and breaks:
SH-3340