For example by uexU9E

VIEWS: 2 PAGES: 12

									   Reallocations Brief for the Statistical Sciences (GSC 14)


                                     January 2002




Submitted by the Steering Committee for the Statistical Sciences:
                 Christopher Field, Dalhousie University
                 Christian Genest, Université Laval
                 Peter Hooper, University of Alberta
                 Jack Kalbfleisch, University of Waterloo
                 John Petkau, University of British Columbia (Chair)
                 Nancy Reid, University of Toronto
                 Bruno Rémillard, École des Hautes Études Commerciales




                                                               December 30, 2001
                                            Summary
The statistical sciences have an essential role to play in the knowledge-based environment and
economy of the 21st century at the centre of which lies the need to collect, summarize and
interpret information in support of science, technology and policy decisions. Canadian statistical
scientists have great expertise in the theoretical core of the subject and in a wide-ranging
spectrum of application areas. They are also key members of multidisciplinary research teams
addressing national issues. Canada is well positioned to develop further this vital discipline
where demand for highly qualified personnel (HQP), both within and outside academia, has been
robust and is increasing at a rapid rate.
Our vision for the statistical sciences in Canada is that of a discipline with broad-based
excellence in all research universities that builds on current outstanding strengths in probability
and statistics. Effective graduate programs will meet the demand for HQP in all, but especially
interdisciplinary, areas. Building on initiatives from the previous reallocations exercise,
researchers in the statistical sciences will expand their activities in several strategic research
areas for Canada: computationally intensive methods, methodology for the health and biological
sciences, risk modeling and assessment, and methodology for economic, social and health policy.
We request an allocation of $1,100K to be directed toward the vision by increasing support to
more competitive levels for our strongest researchers and research proposals, by developing
research capabilities in the key areas identified, and by providing incentives to promote
interdisciplinary activity in graduate programs. This allocation will pay large dividends to
Canada in terms of strong relevant research and help to maintain and attract the best personnel,
both researchers and students, in this strategic discipline.
1. Introduction
The knowledge-based economy of the 21st century is increasingly dependent on the collection
and analysis of data to make projections and informed decisions in the context of uncertainty.
This dependence arises in industry, business, government, health, social policy and planning,
environmental management, and many other areas. Enhanced computing capacity and
communication technology have initiated this revolution, leading to massive amounts of data of
varied and increasingly complex types and the need for information technologies to handle those
data. The concomitant need to present and interpret relevant information accurately and clearly
depends critically on the development and application of statistical methods to optimize the
collection, organization, and analysis of data. This technological revolution has created an
increased need for the statistical sciences to draw valid information from the data.
To support economic, social, medical and technological development, Canada needs a broadly-
based community of statistical scientists who can design efficient data collection mechanisms,
develop the theory and methods of inference required to extract essential features of data beyond
chance variation, and elaborate prediction and decision tools that take into account uncertainty
and risk. There has been a robust and growing demand for such people in government, industry,
business, health and education, and this need and demand will continue to increase.
The need for statistical methods is inextricably linked to advances in science and engineering,
and new challenges continually arise. For example, how can one use magnetic resonance


                                                1
imagery data to help understand the evolution of brain cancer or noisy satellite data to study
global climate changes? How can gene expression data help in identifying genes responsible for
a given disease? How can massive amounts of financial data be more effectively used for
prediction or interpretation? How can we best design complex longitudinal studies to assess use
of the health system and related economic issues? How can extensive data on HIV and AIDS be
used to track the epidemic and how can efficient treatment modalities be determined? Questions
such as these have substantial statistical components and their effective resolution relies on new
and existing knowledge in the statistical sciences as well as in other disciplines.
The Statistical Sciences’ previous reallocations brief identified five emerging areas of statistical
activities of major and immediate societal importance to Canada. These were related to the
analysis and exploitation of massive data sets, environmental science, genetics and medical
science, stochastic modeling of complex systems, and social policy statistics. Initiatives have
since been successfully implemented and strength in these areas is currently being enhanced.
Our vision is presented in Section 2, and the current status of the discipline is summarized in
Section 3. Strategies to strive toward this vision are identified in Section 4, and specific
proposals for reallocated funds are made in Section 5. The balance of the brief describes
consequences of a denial of reallocated funds (Section 6), impacts of the previous reallocations
decisions (Section 7), and other issues (Section 8). Concluding comments are given in Section 9.
2. A Vision for the Statistical Sciences in Canada
Given that data collection, organization, analysis and interpretation are inherent to science and to
the knowledge-based economy, Canada needs a strong and broadly-based community of
statistical scientists to help maintain its status among industrial nations. Our vision is:
A) Research in the statistical sciences will have a broad geographical base of excellence.
Quantitative research that depends on the statistical sciences is pursued in universities, research
centres, government agencies, and industry all across the country. It is thus important to promote
and maintain research capability and graduate training in the field at all research universities.
B) The best researchers will have adequate support to promote work at an internationally
competitive level. The statistical sciences in Canada have a distinguished history and a very high
international profile. The ability to maintain and develop world-class research groups in today’s
competitive environment requires a much larger financial commitment to our best researchers.
C) The research environment will foster collaboration and interdisciplinary involvement that
leads to strong relevant statistical work. This is crucial, as research in the statistical sciences is
stimulated by involvement in other disciplines which, in return, benefit from statistical input.
D) Graduate programs in the statistical sciences will be larger and well supported. The demand
for highly qualified personnel (HQP) in the statistical sciences has traditionally been very solid
in both theoretical and applied areas. Newer domains like bioinformatics and genetics, finance,
data mining, and environmental risk have resulted in a rapid increase in need and opportunity.
E) There will be deep statistical involvement in areas of societal importance. Our previous brief
identified key areas of massive data sets and information technology, genetics and medical
science, stochastic modeling, and environmental science for development. There has been a keen
response that must continue to ensure Canada’s continuing leadership in the field.


                                                  2
F) A national network for research in the statistical sciences will be established. The
accomplishment of this is beyond the scope of this brief. It is, however, an important aspect of
the vision that will pay great dividends to Canada and the discipline.
3. Current Status of the Canadian Statistical Sciences Community
3.1 Status of Canadian Researchers in Statistical Sciences. The Canadian statistical
community is broad and has secured an international reputation for excellence. To assess its
status, we extended earlier surveys (Genest 1997, 1999) of contributions by country and
institution, to publication in the leading 16 statistics and 9 probability journals (Genest 2002).
For the period 1986-2000, Canada ranks second worldwide in statistical research, whether in
terms of total output (behind the US) or on a per capita basis (behind Australia). Among the 100
most prolific authors in statistics, four are from Canadian universities and five are Canadians
working in the US. An impressive 11 Canadian institutions rank among the top 100 (Waterloo,
Toronto, UBC, Carleton, McGill, Montréal, Western, Statistics Canada, Simon Fraser, York,
Laval). In probability, Canada ranks fifth worldwide, exceeded only by the US, France, Israel
and the UK. Adjusting for population, Canada stands eighth. Four Canadian institutions
(Carleton, UBC, Toronto, the Fields Institute) rank among the top 100 worldwide. Three
Canadians are among the 100 most published authors; they actually rank among the top 20.
There are many other indicators of the strength of the community. Canadian statistical scientists
are widely represented on editorial boards of the major journals, are frequent participants and
organizers of international conferences, play active roles in world organizations such as the
Bernoulli Society and the International Statistical Institute, and have been widely recognized
with fellowships in the Institute of Mathematical Statistics and the American Statistical
Association. The international COPSS Award, made annually for outstanding research
contributions by a statistician under the age of forty, has been awarded five times to statisticians
educated in Canada in its 23-year history. The Fisher Lectureship, the leading award to senior
statisticians, has been awarded twice in the 1990s to researchers working in Canada. The
Canadian Journal of Statistics is a first-rate research journal with broad-based appeal that
continues to grow in stature. Statistics Canada, internationally regarded as preeminent among
statistical agencies, relies heavily on a strong contingent of methodologists and emphasizes
research in design and analysis of surveys. Its flagship journal, Survey Methodology, attracts
authors and readers from many countries.
3.2 Areas of Strength in Canada. Since the 1940s, Canadian statistical scientists have made
wide-reaching contributions at the theoretical core of the discipline and in applications motivated
notably by problems in biology, health, agriculture and the social sciences. New research
questions arise through issues encountered in the modern information-based areas of science,
technology, health and business. This leads to an ever-increasing role for computationally
intensive methods and brings the statistical sciences in close contact with other disciplines. This
is reflected by the increased interest and activity in key areas identified in our last brief.
There are outstanding probability groups at Carleton, Ottawa, Toronto, York, UBC and Alberta,
and excellent young researchers at other universities across the country, working on a wide
variety of problems in stochastic modeling. For example, Canadian researchers are world leaders
in the theory of measure-valued diffusions and strong/weak approximations in empirical
processes; they have made important contributions in change-point analysis, the theory of Monte


                                                 3
Carlo and Markov chain Monte Carlo (MCMC) methods, particle methods for non-linear
filtering of high-dimensional data, and the development and simulation of financial models. This
work has wide-ranging applications to fields like finance and insurance, risk management, the
physical, medical and biological sciences, and engineering. For example, researchers at HEC,
Toronto, Waterloo and UBC are developing stochastic models that go beyond the classical
Black-Scholes model for evaluating sophisticated investment instruments such as financial
derivatives, and actuarial scientists at Laval, Toronto, Waterloo, Western and elsewhere are
developing models and approaches for valuing complex insurance portfolios in various lines of
business. Waterloo’s actuarial sciences research group is widely regarded as the best worldwide.
Statistical theory and inference has long been an important area of research strength in Canada
and this activity at the core of the discipline continues to be a major thrust. It draws motivation
from diverse applications and forms the common theoretical fabric from which applications are
addressed; some of these are discussed in more detail below. Canadian researchers continue to
play key roles in the development of likelihood methods of inference and the geometry of
statistical models. Higher-order asymptotics for the likelihood and related improvement of
approximations in finite samples, the theory of estimating functions, robust estimation and
Bayesian methods are current areas of research with wide-ranging theoretical and applied
implications. Toronto, Waterloo and UBC have traditionally been strong in these areas but
researchers from many institutions are contributing in important ways. Each of the broad areas of
experimental design, nonparametric statistics, multivariate analysis and time series are also
strengths at several universities throughout the country.
Canadians have been very influential in the theory and development of survey methodology.
Statisticians at Carleton, Waterloo, Western, Toronto, Laval, Montréal, Alberta, SFU and
Statistics Canada are leaders in research in the foundations of sample survey theory, estimation
for small areas, imputation of missing survey data, re-sampling methods to assess uncertainty for
complex survey designs and nonparametric approaches to the analysis of survey data. Statistics
Canada promotes interaction between its methodologists and academic researchers to great
mutual benefit and hires substantial numbers of graduates in the statistical sciences each year.
Market pressure will be further fuelled by the establishment in Montréal of the UNESCO
Institute for Statistics, which was attracted by Canada’s reputation of excellence in statistics.
Canadians have also spearheaded many developments in computationally intensive methods.
Researchers at McGill, UQAM and UBC have developed innovative methods for nonparametric
regression, smoothing and curve estimation that are widely applicable for data exploration,
pattern recognition, model fitting, and decision-making based on large datasets. Researchers at
Montréal, Carleton, Waterloo, McMaster and SFU have also made significant contributions to
re-sampling techniques such as the bootstrap; others at McGill, Ottawa, Toronto, Western and
UBC have developed MCMC methods for the analysis of complex stochastic models. New
collaborations with computer scientists, notably at Montréal, Waterloo and UBC, have also
resulted in important advances to data mining methodology.
Canadians have also contributed greatly to methodology for the health sciences. Researchers at
Waterloo are world leaders in survival and event-history analysis, widely used in modeling and
analyzing disease histories for persons with HIV/AIDS and other conditions. There is substantial
strength in methods for analyzing positron-emission tomography and magnetic resonance
medical imaging data at UQAM and McGill. Canadian biostatistical scientists have also made


                                                4
significant contributions to sequential designs for clinical experiments and to methods for
analyzing response-sensitive sampling schemes, for dealing with missing data in longitudinal and
survey studies, and for disease incidence and surveillance, among other topics.
This outstanding base in biostatistical research is the foundation for growing involvement in
methodology for the biological sciences, especially in genetics, genomics and bioinformatics.
With molecular biologists, statisticians at Dalhousie are addressing issues associated with
building and evaluating phylogenetic trees that accurately reflect genome evolution. Researchers
at UBC, Alberta, and Toronto are addressing the computational and inferential issues that arise
with micro-array and gene expression data and are developing innovative methods for predicting
genetic and protein structure based on nucleotide or amino acid sequences. With human disease
scientists, researchers at McGill, Toronto, Guelph and UBC are developing flexible statistical
models to examine interactions between genetic variation and environmental exposure on disease
status. Other developments include, at SFU, likelihood techniques to assess genetic variation
among populations and, at Dalhousie and Laval, models incorporating unknown pedigree
relationships to map quantitative trait loci in natural tree populations.
Another area of broad strength is methodology for the environmental sciences. Canada has long
been a leader in statistical studies of wildlife and ecology. Researchers at SFU, Laval and in the
Atlantic provinces are playing important roles in resource management and environmental
impact assessments by developing approaches to surveys of wildlife and fisheries and models for
the analysis of the collected data. Statistical scientists at UBC, Toronto, Ottawa and McGill are
developing methodology for environmental epidemiology, notably to assess health risks
associated with air pollutants; others at SFU have developed spatial methods for disease mapping
to assess associations with exposures. Researchers at Dalhousie, Alberta, UBC and the Canadian
Climate Centre have also made significant contributions to methods for designing monitoring
networks and to spatial-temporal modeling of natural resources and global climate change.
Finally, the Institute for Improvement in Quality and Productivity at Waterloo is widely
recognized as a leading university-based research group in industrial statistics, and groups at
Manitoba and SFU have also developed synergistic linkages with industrial research. Areas of
strength include the design and analysis of experiments for product and process improvement,
large-scale computer experimentation, multivariate process control, and reliability analysis.
3.3 HQP, Discipline Dynamics and Research Support. As public reports do not always
distinguish between the mathematical and statistical sciences, a survey of Canadian universities
was conducted in June 2001. Although all 40 units with at least three tenure-track faculty in the
statistical sciences responded, these data somewhat underestimate the total Canadian activity.
These 40 units reported about 540 graduate students enrolled in their programs. Only 8% of these
students held NSERC Scholarships, while another 14% held other major scholarships from
external sources. There was a relatively small pool of postdoctoral fellows – about 30 per year
nationwide. Many of these positions were of short duration, taken until an appropriate time to
enter the job market; others were held by graduates from other disciplines seeking to develop
expertise in biostatistics, actuarial science, finance, and other specialties of high demand.
The 40 units had 592 Masters and 120 Ph.D. graduates in the 4-year period 1997-2001. The
number of Masters graduates has increased steadily from about 80 per year in the early 1990s to
the present level (an 80% increase). The rate of Ph.D. graduations remained relatively constant


                                                5
over this period due to an exceptionally attractive job market for Masters graduates and keen
international competition for Ph.D. students. Of Masters graduates taking employment, 55%
went to industry, 19% to government agencies and 26% to universities or research groups. Of
Ph.D. graduates, 60% took up academic posts (more took up tenure-track positions than
postdoctoral fellowships), 20% went to industry and 20% to government agencies or research
groups. Of Ph.D. graduates, 68% took their first employment in Canada.
The number of full-time equivalent tenure-track faculty (about 300) remained almost constant
over the past four years. There was a 20% turnover in these positions, about half due to
retirements and half to resignations. Over the next two years, 48 new tenure-track openings are
expected (including 19 Canada Research Chairs), but only about 20 retirements. These figures
exclude the 23 reported current vacancies and all contractually-limited positions. With only 30
Ph.D. graduates per year, there will be a serious shortfall in filling these posts, especially with
fierce international competition and ever-increasing demand for Ph.D. graduates from the non-
academic sector. There is a pressing need to increase the number of Ph.D. graduates.
Among statistical scientists in these 40 units, annual funding from NSERC Individual Research
grants totaled 4.3M in 2000-1, or 42% of their 10.3M funding directly targeted for research in
the discipline. Reported total funding increased from 8.0M in 1997-8 to 10.3M in 2000-1. The
actual funding is much larger as only portions of large grants, for example, in health, behavioural
and genomics projects, are included. NSERC Individual Research grants are especially important
in supporting curiosity-driven research and in leveraging other research support.
Statistical scientists are involved with various other programs and organizations that support
research in Canada, most notably the three Mathematical Sciences Institutes (CRM, Fields and
PIMS). The CRM held a theme year in statistics in 1997-8, and there were programs in
probability and on graphical models in statistics at Fields in 1998-9 and 1999-2000, respectively.
The Institutes have also provided some funding for annual meetings of the Statistical Society of
Canada, Pacific Northwest Statistics Days, and some regional workshops. The Carleton-Ottawa
Laboratory for Research in Statistics and Probability, supported by an NSERC MFA grant, also
sponsors many activities. The MITACS Network of Centres of Excellence has played an
important role in bringing together and providing support for statistical scientists in genetics,
data mining, industrial mathematics, probability and finance. The Canadian Institutes of Health
Research is also a development of substantial interest; statisticians are involved in its review
panels and it is becoming an important source of research funds. Similarly, programs like the
FCAR Team Research Grants in Québec and the Premiers Research Excellence Awards in
Ontario provide research support to outstanding researchers. The existing and expected openings
for Canada Research Chairs will help to keep and attract outstanding statistical scientists in a
very competitive market. Various infrastructure programs such as the CFI and provincial
initiatives have also been important sources of support, though lack of funds for maintenance and
technical needs is often a limiting factor. There are also sources for postdoctoral support from
MITACS and the Mathematical Sciences Institutes that can help enable graduates from other
disciplines to seek postdoctoral training in key areas of the statistical sciences.
4. Strategy
The following strategy is designed to promote strength in theory and methods in the statistical
sciences, to build formidable interdisciplinary activity at Canadian universities, and to support


                                                6
and promote development in areas of important societal and scientific application. The divisions
below refer to elements of the vision and offer specific strategies for attaining them.
A) Geography and infrastructure. Because the statistical sciences interact with many other areas,
it is vital to maintain strength, research activity and a training component at all research
universities. It is also crucial to provide good research support to new researchers to accelerate
their development and to be competitive in a strong international market. These have been and
will continue to be high priorities of Grant Selection Committee (GSC) 14.
B) Support for the best researchers. The last NSERC Reallocations Committee wrote: “The
Canadian Statistical Sciences community is strong and its members compare well at the
international level. (…) The field … is very broad in its applications despite a relatively low
level of funding and a small community.” Despite this recognition of quality and a 42% increase
in average grant since 1992 (2nd largest among GSC’s), GSC 14 still has the lowest average grant
($16,570/year; March 2001 Discipline Dynamics Report, Table 3). At the same time, there are
increasing demands both within academia and outside for HQP. Given the need to provide a high
initial grant to the many new (both junior and senior) researchers in the discipline, there is
intense pressure on grant levels for other applicants to GSC 14. It is important to ensure that
strong researchers in high demand areas have sufficient funds to provide appropriate support for
graduate students and hire essential technical personnel. Therefore, increased funding is sought
for statistical scientists with excellent research programs. This will help to attract and retain these
individuals in Canada, and enhance graduate training.
C) Foster collaborative and interdisciplinary work. There is an essential interplay between
theory, methods and applications in the statistical sciences. New areas of investigation constantly
evolve through expanding and changing needs of researchers in science, engineering, the social
sciences, health, etc. There is also a large demand for graduates with a solid theoretical base
combined with experience in interdisciplinary collaboration. Accordingly, outstanding statistical
scientists involved in interdisciplinary work should be provided with additional support to
enhance that research through the increased involvement of students. As an important ancillary
benefit, additional statistical expertise will come to bear on research in other disciplines.
D) Develop and support graduate programs. This element is addressed in part by B) and C), but
also by E) below, which is designed to support research in key areas of societal importance.
E) Promote statistical involvement in key areas. The main thrust of the previous brief was
successful in developing increased interest among statistical scientists in areas of major
importance to Canada. It is imperative to renew our commitment to this thrust and the key areas
are described below. Effective work in all of these areas requires substantial support in technical
programming as well as up-to-date hardware and software environments. These are also areas of
high demand for HQP and competitive research support for graduate students is essential.
I- Computationally intensive methods. Additional investment is required to address challenges
arising from the continuing explosion of data collected for scientific investigation, decision-
making and public policy. Examples include transactional data, telecommunication and internet
traffic flow data, and data for monitoring individual-level health care use or environmental issues
such as energy consumption. Collaboration with computer scientists in database design, artificial
intelligence and graphics, must be combined with development of complex stochastic models
and new methods of statistical analysis. The amount and potentially uneven quality of the data


                                                  7
are major considerations. Robust methods, new smoothing and functional data analysis
techniques, and stochastic models that reflect spatial-temporal components are essential. The
complexity of the models and data structures will require the use of re-sampling, MCMC and
other computationally intensive approaches for statistical inference.
II- Methodology for health and biological sciences. Development of biostatistical methodology
is essential for leading-edge health research and sophisticated new statistical methodology is
required to effectively extract information in genetics, genomics and bioinformatics. There are
major computational aspects to many problems in the latter areas, but statistical scientists are
uniquely qualified to develop the complex stochastic models and methods required to address
many of the important scientific questions. In collaboration with subject-area specialists,
researchers are making substantial improvements to the methodology available. Further
investment in this key area will build on these exciting developments.
III- Risk modeling and assessment. The foundation of research in this area is stochastic models to
capture and help interpret characteristics of the phenomena under study. In actuarial science,
interest centres on the management of complex, interdependent insurance portfolios. In finance,
challenging research arises in valuating complex new financial instruments and schemes in
rapidly changing markets. The development of realistic stochastic models requires sophisticated
probability theory and checking their adequacy often requires new statistical tools. Stochastic
models in environmental and health risk assessment must incorporate the spatial aspects and the
subtlety of effects being measured demands careful modeling and efficient methods of inference.
Canadian statistical scientists have made important contributions in this highly interdisciplinary
area and further investment will build on this strength.
IV- Methodology for economic, social and health policy. The great strength of Canadian
statistical scientists in survey methodology forms a foundation for development of much needed
new methodology. Procedures tend to be computationally intensive, as when assessing
uncertainty in complex surveys through re-sampling or using neural networks for data editing.
Sampling using time-series methods and small-area estimation using MCMC and hierarchical
Bayesian methods are becoming important. Economic and social planning make increasing use
of statistical models of populations over time and an emerging area of major importance is the
design and analysis of longitudinal surveys; examples are provided by the National Longitudinal
Survey of Children and Youth, or the Survey of Labour and Income Dynamics. These trends
raise challenging statistical problems combining complex survey designs and longitudinal data
analysis. There is a unique opportunity to build collaborations between these two areas of
substantial Canadian expertise. The Canadian Initiative in Social Statistics, co-sponsored by
SSHRC and Statistics Canada to strengthen capacity in quantitative social science research,
provides further impetus.
5. Specific Proposals
The total Research Grants budget for GSC 14 is $4,552K, 10% of which is contributed to the
reallocation pool. An allocation from the pool of $1,100K is requested for three proposals,
presented here in order of priority.
Proposal 1: Increase funding to the best researchers ($360K). Senior statistical scientists in
Canada are strong by any standard. Yet, their current funding levels do not allow adequate
discretion in supporting graduate students, postdoctoral fellows or technical support personnel.


                                                8
Funding pressures are also especially severe for promising young researchers seeking first or
second renewals. With an average renewal grant of only $17,500/year and selectivity to fund
senior applicants, it is very difficult to provide these researchers with increases that can support
greater involvement with HQP, international contacts, and access to technical support. The
requested $360K would ease these difficulties by providing extra allocations to the research
grants of perhaps 30 of the strongest statistical scientists with excellent research proposals.
Proposal 2: Support research in key emerging and developing areas ($540K). Section 4 outlined
directions for development of four key areas, each of which is a crucial direction for statistical
research. The allocation of $540K will provide additional support to about 45 researchers to help
meet needs for computing support and to provide funding for graduate students and postdoctoral
fellows in these areas where the demand and need for HQP are particularly intense.
I- Computationally intensive methods. This rapidly evolving area presents exciting challenges
arising from the volume and complex structure of modern data. Canadians have been at the
forefront of many developments in this area and it is crucial to build on this strength. Funding
levels similar to those in Computing and Information Sciences are needed to sustain growth.
II- Methodology for health and biological sciences. This is a hot area with much Canadian
strength though the competition for our best researchers and graduates, notably from the US, is
severe. The special allocations will provide the strongest Canadian researchers with more
competitive funding and expand interactions with the subject-area research communities.
III- Risk modeling and assessment. This interdisciplinary key area permeates all aspects of
modern society. Statistical science is vital in designing studies to assess risk, in weighing
evidence on cause and effect, and in communicating findings to policy makers and the public.
Particular areas include financial, actuarial, environmental and epidemiological risk.
IV- Methodology for economic, social and health policy. New methodology for the design and
analysis of complex longitudinal surveys is essential in many areas including government,
market research, health and the social sciences. This key area will build collaborations between
existing strengths in longitudinal studies and complex survey designs.
Proposal 3: Support interdisciplinary research ($200K). As part of the Research Grants
Program, the GSC will allocate “Interdisciplinary Research supplements” of about $10K/year to
outstanding statistical scientists involved in work in such areas. These allocations will be made
in response to projects, documented in the research proposal, to involve and support students and
postdoctoral fellows in interdisciplinary teams. These projects will entail direct contact with
statistical problems in other disciplines, ideally through interdisciplinary supervision.
6. Consequences of No Reallocated Funds
While the Statistical Sciences community is currently funded at a low level, it carries out a broad
and successful research program. If reallocated funds are denied, the GSC’s highest priority
should continue to be the provision of reasonable starter grants for new and young researchers, as
this is crucial to the future. However, it would soon become very difficult to maintain activity in
computer-intensive research or in areas requiring support personnel. Statistical scientists would
also be unable to collaborate fully in interdisciplinary research and there would be even less
support for graduate students at a time of increasing demand for HQP from the public and private


                                                 9
sectors, leading to reduced research capacity. Most significantly, there would be great concern in
the community about a lack of financial and moral support that would have widespread effects,
including a potential increase in emigration among the best students and researchers.
7. Impact and Consequences of the Implementation of the 1998 Reallocations Decision
Proposals made in the previous submission sought to expand support of research in key and
emerging areas of importance to Canada. The strategy was to increase funding to researchers of
exceptional promise to integrate additional students, postdoctoral fellows and research associates
into these areas, thereby enhancing research capacity. An allocation of $560K was secured for
four areas: massive data sets and information technology, genetics and medical science,
stochastic modeling, and environmental science. The reallocations brief was very effective in
directing and enhancing interest in these key areas, to which roughly 60% of funded proposals,
both news and renewals, were related in subsequent competitions. Reallocated funds, which were
designated only to proposals of excellent quality, helped to direct a vital and growing research
effort toward particular areas of importance in the discipline and resulted in a targeted allocation
of support to students and to high quality research. The reallocated funds also aided in generating
other funds by promoting involvement of statistical scientists in genetics, data mining and
industrial statistics, all priority research areas of the MITACS Network of Centres of Excellence.
8. Other Issues
The development of a national network for research in the statistical sciences is an important
aspect of our vision that will be vigorously pursued over the next four years. Its purpose will be
to draw together a geographically diffuse community, aid in the development of links between
universities, industry and the public sector, and promote statistical research in areas of strategic
national importance at Canadian universities and other institutions. In collaboration with the
Mathematical Sciences Institutes, this network will serve to coordinate statistical activities
nationwide. The interdisciplinary proposal with the Institutes for a multi-year National Program
on Complex Data Structures is related to the key area of methodology for economic, social and
health policy and will help develop the basis and linkages needed for such a national network.
9. Conclusion
Canadian statistical scientists have an enviable international reputation for accomplishment in
basic and applied research. This brief has outlined key areas that will continue to drive much of
the development in the discipline over the next few years. Research in these areas is highly
computational and the cost of research is comparable to fields such as industrial engineering and
computer science, where grants are much higher on average. There is a strong and fast-growing
demand for HQP with both Masters and Ph.D. degrees in key areas of statistical research. There
is also a need for postdoctoral training to help develop (and sometimes redirect) research
programs of new graduates. The need for increased research support is very real.
Our reallocation request is $1,100K. This net increase of $645K, or 14% over the current annual
budget for research grants in the statistical sciences, is not a large reallocation in terms of dollars.
Yet it will greatly strengthen the research capacity of statistical scientists in Canada. It will pay
large dividends to Canada in terms of strong relevant research, and in maintaining and attracting
the best personnel, both researchers and students, in a discipline of strategic importance.



                                                  10
References
Genest, C. (1997). Statistics on statistics: Worldwide performance based on journal publication
over the period 1985-1995. The Canadian Journal of Statistics, 25, 427-443.
Genest, C. (1999). Probability and statistics: A tale of two worlds? The Canadian Journal of
Statistics, 27, 421-444.
Genest, C. (2002). Worldwide research output in probability and statistics: An update. The
Canadian Journal of Statistics, 30, in press.




                                              11

								
To top