National Program in Complex Data Structures

  • Start Date: 04/24/2008
  • End Date: 04/25/2008

Dr. Igor Burstyn.
Dr. Mahyar Etminan.
Dr. Adrian Levy. Centre for Health Evaluation and Outcome
Dr. Bill Leslie.
Dr. Malcolm Maclure.
Dr.Ed Mills.
Dr. Carl Phillips.
Dr. Jat Sandhu.


University of British Columbia


Dr. Igor Burstyn.
Occupational Medicine, University of Alberta.
Title: A vignette from occupational epidemiology: Stitching evidence from tattered fabric

Dr. Mahyar Etminan.
Centre for Clinical Epidemiology & Evaluation, Vancouver.
Title: TBA

Dr. Adrian Levy. Centre for Health Evaluation and Outcome Sciences;
Department of Health Care and Epidemiology, UBC.
Title: TBA

Dr. Bill Leslie.
Department of Internal Medicine, University of Manitoba.
Title: TBA

Dr. Malcolm Maclure.
Pharmaceutical Services Division, BC Ministry of Health.
Title: You randomize. We Analyze.

Dr.Ed Mills.
Faculty of Health Sciences, SFU.
Title: TBA

Dr. Carl Phillips.
Department of Public Health Sciences, University of Alberta.
Title: Can quantitative methods help detect and reduce "publication bias in situ"?

Dr. Jat Sandhu.
Vancouver Coastal Health.
Title: TBA




The National Institute for Complex Data Structures (NICDS), via its project on "Statistical Innovation for the Analysis of Complex Data in Medical and Health Sciences" is pleased to announce two forthcoming events.

These two related but distinct events will take place April 24-25, 2008, in Vancouver, at the Pacific Institute for the Mathematical Sciences. Interested participants are encouraged to register for either or both events. (Registration is free, but required.)

The first event is Understanding and Managing Mismeasured Variables in Biostatistical Analysis, a short course being held on Thursday, April 24th. The second event is Methodological Needs and Desires in Public and Population Health Research, a workshop being held on Friday, April 25th.





Financial Support

Funding is available to assist graduate students and research trainees from outside Vancouver who wish to attend either or both events. Please contact Paul Gustafson ( for details.


Short Course (Thursday April 24)

"Understanding and Managing Mismeasured Variables in Biostatistical Analysis"

Regression techniques are the central statistical tools for
investigating the relationship between an outcome variable and a set of
explanatory variables. A common concern in many subject-areas is how to
carry out and interpret regression modelling when one or more of the
explanatory variables is poorly measured. This is particularly of
concern in biostatistical contexts. In epidemiology, for instance,
there is often exposure measurement error or misclassification when
exposure-disease relationships are being inferred. Whereas there is
some general awareness that unchecked measurement error in explanatory
variables can produce misleading inferences, often there is a lack of
action to mitigate the problem. Other problems (e.g. missing data)
often seem to garner more attention.

This one-day course will survey current thinking and methodology
surrounding statistical inference when explanatory variables are
subject to measurement error or misclassification. Some emphasis will
be given to exploring the deleterious consequences of implementing and
interpreting regression analysis as if the measurement error were
absent, and more emphasis will be given to modelling strategies which
explicitly account for measurement error.

Various types of measurement error and misclassification will be
introduced and distinguished (non-Berkson versus Berkson,
nondifferential versus differential), and their impacts if ignored will
be discussed. Various modelling methodologies will be explored,
including likelihood and Bayesian techniques, regression calibration,
and simulation-extrapolation. The strengths of assumptions required for
the methods will be examined, and the required nature and strength of
information about the measurement error magnitude will be addressed.

The course will be organized into six lectures. In the three morning
lectures, the three instructors will introduce the main ideas and
techniques. The three afternoon lectures will focus on more specialized
topics of current relevance and interest.

To make the course relevant to as wide an audience as possible, the
emphasis will be on concepts and modelling issues, rather than
technical and mathematical issues. There is no prerequisite for these
lectures. Basic knowledge of regression should be sufficient to
understand most of the material covered. Software and implementation
issues will be discussed.

Workshop (Friday April 25)

"Methodological Needs and Desires in Public and Population Health Research"
The workshop "Methodological Needs and Desires in Public and Population
Health Research" will feature talks by health researchers, describing
the quantitative methodological challenges they face, and their
wish-lists for innovation from the statistical community. It is hoped
that mixed workshop participation of both health researchers and
statisticians will help foster technology transfer of innovative
statistical methodology into health research applications.



The 2008 National Program in Complex Data Structures Committee
Professor Paul Gustafson (, Department of Statistics, University of British Columbia

Other Information: 


To register, please click here.