14.10.2013 Views

Annual Report 2005 - Fields Institute - University of Toronto

Annual Report 2005 - Fields Institute - University of Toronto

Annual Report 2005 - Fields Institute - University of Toronto

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

Career: Some Remarks. On August 7, Dick DeVeaux (Williams<br />

College) addressed teaching statistics in an invited<br />

speech on Math is Music; Statistics is Literature – or Why are<br />

there no 6 year old novelists? In addition, two invited panel<br />

sessions featured candid remarks and advice from journal<br />

editors (Annals <strong>of</strong> Statistics, Canadian Journal <strong>of</strong> Statistics,<br />

Technometrics and Biometrics) for publishing, and grant<br />

directors/managers (NSF, NSERC, NIH, ONR) for grant<br />

applications.<br />

One highlight <strong>of</strong> the 2004 NRC is the first time nomination<br />

<strong>of</strong> the Richard L. Tweedie New Researcher Award. The<br />

award recipient will receive a prize <strong>of</strong> up to $2000 and be<br />

invited to give a special lecture at the NRC. The IMS Committee<br />

on Special Lectures will decide the award winner.<br />

In summary, the 2004 NRC was a great success, <strong>of</strong>fering<br />

a comfortable setting for new researchers to share their<br />

research and make connections with their peers in informal<br />

settings such as short research presentations and various<br />

social programs.<br />

Special Addresses:<br />

Richard D. DeVeaux (Williams College)<br />

Math is music; statistics is literature – or why are there no 6<br />

year old novelists?<br />

Xihong Lin (Michigan)<br />

Exploring roads to a successful career: some remarks<br />

Terry Speed (UC Berkeley)<br />

How to do statistical research<br />

Jeff C.F. Wu (Georgia Inst. <strong>of</strong> Tech)<br />

Statistics and statisticians: an amateur’s tour guide<br />

Workshop on Missing Data Problems<br />

August 5–6, 2004<br />

Held at the <strong>Fields</strong> <strong>Institute</strong><br />

Organizers: Richard J. Cook and Don L. McLeish (Waterloo)<br />

Modern data <strong>of</strong>ten includes some form <strong>of</strong> censorship or<br />

missing data. Data imputation is a critical component <strong>of</strong><br />

the analysis <strong>of</strong> such data and crude methods for data imputation<br />

can lead to substantial bias in the results and the<br />

conclusions. Missing data problems are common in health<br />

research (e.g. retrospective and prospective studies), sample<br />

surveys (e.g. non-response), and less obvious parts <strong>of</strong> any<br />

study in which the data available is influenced by what is<br />

G e n e r a l S c i e n t i f i c A c t i v i t i e s<br />

Jamie Stafford, Richard Cook and Don McLeisch<br />

easy or feasible to collect. Longitudinal studies which collect<br />

data on a set <strong>of</strong> subjects repeatedly over time are subject<br />

to attrition, subjects drop out because they move and<br />

suffer side effects from drugs, or for other <strong>of</strong>ten unknown<br />

reasons. Similarly in sampling, survey non-respondents are<br />

<strong>of</strong>ten ignored, although factors related to the objectives <strong>of</strong><br />

the study such as income may influence the completeness <strong>of</strong><br />

a subject’s response.<br />

The primary goal <strong>of</strong> this workshop was to provide impetus<br />

to the development <strong>of</strong> mathematical and statistical tools for<br />

the analysis <strong>of</strong> data under various patterns <strong>of</strong> censorship<br />

and mechanisms governing missingness and data imputation.<br />

The workshop brought together researchers from<br />

around the world with common interests in missing data<br />

problems and a broad range <strong>of</strong> approaches for dealing with<br />

them. Approaches ranged from those based on multiple<br />

imputation, inverse probability weighted and more general<br />

classes <strong>of</strong> estimating functions, and EM or generalized EM<br />

algorithms. Issues receiving lively discussion included the<br />

importance <strong>of</strong> robustness, efficiency, identifiability, and the<br />

role <strong>of</strong> sensitivity analyses. The range <strong>of</strong> contexts in which<br />

missing data problems were considered included studies<br />

involving retrospective observation, prospective longitudinal<br />

studies, event history studies, survey sampling, finance,<br />

and social sciences. The meeting was supported by <strong>Fields</strong>,<br />

NPCDS, the <strong>University</strong> <strong>of</strong> Waterloo, and GlaxoSmithKline.<br />

<strong>Fields</strong> <strong>Institute</strong> <strong>2005</strong> ANNUAL REPORT 59

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!