Research interests
My faculty research directory entry
My research interests lie in three separate but inter-related areas. They are
A fourth academic area in which I have
an interest in is that of statistics, science and public policy.>
I have written a few essays on these general topics.
These and other
selected papers and talks are made available below.
Graduate Students
I have been fortunate in being able to supervise a number of students in their
research over the years. Information on these students, including the topic of their
work and in some cases the work itself, may be found
here.
- The quantitative programming environment
Quail (QUantitative
Analisis In Lisp), a Common Lisp extension. Both display-oriented and
command based, Quail is an experimental system for data visualization
and exploration.
- Eikosogram java application
written by Glenn Lee (Master's student and research assistant).
It is based on a series of three papers I wrote in 2003.
- Code and datasets which I have produced
(in S/R, and Quail) for a variety of courses.
Selected papers and talks
Discussion of Per Andersen's ``Goodness of fit in event history analysis using pseudo-observations'' as presented at the Jack Kalbfleisch and Jerry Lawless Conference in May, 2010 at the University of Waterloo. This is a QuickTime movie of my presentation. NOTE: This could take some time to load (discussion.mov is approx 20 MB) and depende on Per's paper for understanding.
Presentation Graphics An important part of data visualization is the development of
effective presentation graphics (as opposed to analytic or exploratory graphics).
This is the principal focus of Tufte's work.
Here are some presentation graphics we developed to communicate the extremely small
probabilities associated with state lotteries.
Figures here are based on Canadian Lotteries but could easily be adapted for others.
-
Warning Labels for Lotteries: Visualizing Small Probabilities
available as a pdf
Poster presentation at the Statistical Society of Canada meeting (Ottawa, 2008) with Rachel Dean.
- The individual "warning labels" themselves are available below.
Please let me know if you would like to use them (in print or electronically).
- Using comparative length to show the probability of winning a major lottery:
Baseball on the train track
- Using comparative areas to show the probability of winning several different major lotteries:
Coins on a football field
- Visualization of expected returns of common lotteries (and video display terminals):
You win some, you lose some (expect to win less than you pay)
- Using length as a time scale to show the expected number of tickets purchased before winning the grand prize for the first time:
When should your ancestors have started buying tickets, for you to have won by now?
- The natural logarithm is often used in statistics to display large or small numbers.
This uses the mathematical constant e as a base, a number that is virtually unknown outside the mathematically trained. The common logarithm uses a base of 10, so that on this scale a unit increase
corresponds to a 10 fold increase on the original scale.
Unfortunately 10 fold increases (or decreases) are also hard to get your head around, especially
for etremely small numbers that are probabilities.
A more ``natural logarithm'' for small probabilities use 1/2 (or possibly 1/6) as base.
A number on this scale corresponds to the number of coins (or dice) which must be tossed simultaneously and all land heads up (or six up).
This poster compares probabilities of a number of familiar events
(Canadian statistics where appropriate) on this scale
to give some indication of the tiny size of the probability of winning the grand prize in
Lotto 649
Applications of Graph Theory to Data Visualization This is joint work with Catherine Hurley of the
National University of Ireland, Maynooth. Best viewed in chronological order.
Picturing Probability or
Eikosograms
Clustering
- A density cluster tree approach to a statistical theory of clustering.
This actively ongoing work which is joint with Wu Zhou (presently my Ph.D. student).
- "Structure of Interactive Cluster Analysis"
- "Interactive Clustering: Overview and Tools"
Keynote talk delivered at
CASI 2001,
the annual meeting of the
Irish Statistical Association,
May 17, 2001. (Slides only; talk included interactive demo)
- See also research essays and theses by my
graduate students on
- "Model based clustering" by Steve Shi
- "A review and implementation of some approaches to metric clustering" by Wu Zhou
- "Clustering of Web Documents" by Rayan Yahfoufi
- "Visual Empirical Regions of Influence (VERI) Clustering: Assessment and Alternatives"
by Erin McLeish
Quail See also the Quail website
- "Statistical Graphics in Quail: An Overview" by C.B. Hurley and R.W. Oldford.
This is the text of an invited paper presented at the 1999 Biennial Meeting of the International
Statistical Institute in Helsinki Finland.
- Quail overview The following three papers go together to give an overview of Quail.
They are three invited papers given over three successive Interface meetings
and show Quail's design and use in three different lights:
- Underlying general principles particularly as they relate to interface:
"Mental Models and Interactive Statistics: Design Principles"
An invited paper presented at the 31st Symposium on the Interface of Computing Science
and Statistics (Models, Predictions, and Computing), Schaumberg, Illinois - June 1999.
This paper was invited for a special session sponsored by the International Association
for Statistical Computing on the subject of User Interfaces for statistical systems.
The organizer's intention was to compare and contrast Display-oriented GUIs with
Command line interfaces.
- Overview and some specific principles:
"The Quail Project: A Current Overview"
An invited paper presented at the 30th Symposium on the Interface, Minneapolis MN, 1998.
The session in which this paper appeared was on Free Software.
-
Use of Quail in teaching:
"Computational Thinking for Statisticians:
Training by Implementing Statistical Strategy"
An invited paper presented at the 29th Symposium on the Interface, Houston TX, 1997.
The session in which this paper appeared was entitled "Educational Issues and Examples".
This covers wide variety of topics. The essays and talks here are
based on invited presentations at the series of small invitation only
conferences developed by Prof. A. Herzberg in honour of her father the
nobel prize (Physics) winner Dr. Gerhard Herzberg. The conferences
are on Statistics, Science and Public Policy and have, for the most part,
been held at Queen's University's Herstmonceux Castle in Hailsham England.
Each year had a different theme and speakers were asked to address
particular subjects in keeping with the theme.
-
J.C. Bailar, III, H.B. Dinsdale, A.M. Herzberg, K.W. James, and R.W. Oldford (2004).
``Statistics, Science and Public Policy: Recommendations for Government and the Scientific Community''. With foreward by the Rt. Hon. Peter Milliken, Speaker of the House of Commons (Canada).
A copy of the report is available in
pdf
and in
html
form.
-
April 2002: Theme was Globalization, Health and the Environment.
I was asked to speak in a session entitled "Globalization: Past and Future".
For this,
I prepared and presented and essay called
"Globalization and Mythology".
-
April 2000: Theme was Society, Science and Education.
I was asked to speak in a session entitled "The University: Past and Future".
For this,
I prepared and presented and essay called
"The University: Past and Future. What's New?".
This is the actual text of the talk as presented:
- April 1999: Theme was The Two Cultures?
which was held in honour of the fortieth
anniversary of C.P. Snow's famous Rede lecture entitled The Two Cultures.
This presentation was made in a session entitled ``The University
Culture''.
-
The presentation I gave was called
"Shifting Cultures: Stimulus to the individual" and
was meant to work in harmony with a presentation by Prof. Mary Thompson of
Waterloo.
Here is the actual text as presented
-
A tidier written version appears in the departmental working paper series of
1999
as ``Statistics, Science and Society: Culture shifts'' Working paper 1999-06 by
R.W. Oldford and
M.E. Thompson
which includes the paper presented by Mary Thompson as well.
My piece is entitled "Cultural Shifts: Humanities to science to
computation":
Each of these appear as a separate chapter in the series of books
Statistics, Science, and Public Policy which may be purchased
by contacting Prof. A. Herzberg of Queen's University in Kingston, Canada.
-
"Scientific Method, Statistical Method, and the Speed of Light'' Departmental
Working paper 2000-02, with R.J. Mackay (~40 pages). He
- "Method, not methods: Teaching a theory of applied statistics" , a talk given at the Department
of Statistical and Actuarial Science at the University of Western Ontario, in Feb. 2008. This is on the novel aspects of STAT231. It can be viewed as
-
"A Note on High Breakdown Regression Estimators" by R.W. Oldford, 1983.
Tech report 39
from the 1983 Center for Computational Research in Economics and Management
Science, MIT,
series. This paper is referenced by Rousseeuuw's JASA paper on Least Median
Square
regression which has been reprinted in a recent volume ``Breakthroughs in
Statistics''.
-
Videos
Some of my early work (1985 - 1988)
was recorded on video and is available for viewing
from the
Video Library
of the
Statistical Graphics Section
of the
American Statistical Association.
Included here are the videos:
-
"Data Analysis Networks in DINDE"
with Steve Peters, 1986.
Here the history of a statistical analysis was represented visually
and in software as a collection of interactive
directed acyclic graphs whose nodes
were steps in the analysis. An edited version of this video was refereed and
appeared in the video publications of CHI-87, the graphical interface/computer
human interface conference in Toronto, 1987. A copy is available in a variety of formats
from the Open Video Project.
The smallest (106.2MB) in mpeg-4 format is available
here.
-
"Hierarchical Views of Statistical Objects"
with Catherine Hurley, 1988.
This is a video where Catherine and I present our design for statistical
graphics.
These ultimately formed the foundation of the graphics and interface
building blocks presently found in
Quail.
-
"Graphical Programming"
with Guy Desvignes, 1988.
This is an early appearance of the idea of graphical programming in a statistical
system. Here the interactive networks of DINDE were used to allow the analyst
the opportunity to capture certain patterns in his/her analysis by pointing
and clicking at an existant analysis. The result would be a new program
automatically created which would perform the same analysisi steps on a
new dataset.
- Pre-university
- A talk given to Grade 10 students visiting the University of Waterloo as part of the ``Waterloo Unlimited'' outreach to high school students. The week long session was on vision, and this talk
was Escaping Flatland: Visualizing Higher Dimensional Data. It is a clickable Quicktime movie.
- Pre-university mathematics on the web. From the UW Math Faculty's
Grains of Truth
July-December 2001.
- University
- "Method, not methods: Teaching a theory of applied statistics" , a talk given at the Department
of Statistical and Actuarial Science at the University of Western Ontario, in Feb. 2008. This is on the novel aspects of STAT231. It can be viewed as
- Physical laboratories. Six were developed for teaching statistical concepts in an
introductory course by myself and Jock Mackay. In nearly every one, physical
equipment was designed and built (about 75 copies each). They were:
- Eikosogram java application
written by Glenn Lee (Master's student).
This is a graphical tool for teaching probability and such concepts as
conditional probability, Bayes rule, and conditional independence.
It can also be used to display categorical data (related to Mosaic plots).
It is based on a series of three papers I wrote in 2003.
See
above
for these three and a talk on eikosograms vs Venn diagrams
-
The University Past and Future: What's New?
(pdf format).
May 2000.
-
Cultural Shifts: Humanities to Science to Computation (pdf format).
July 1999.
-
Computational Thinking for Statisticians: Training by Implementing
Statistical Strategy
(pdf version) 1997.
Sept. 2008.
rwoldford (at) uwaterloo.ca