|
Thomas M. Loughin Professor Department of Statistics and Actuarial Science MAILING ADDRESS: Statistics and Actuarial Science ¶¡ÏãÔ°AV 8888 University Drive Burnaby, BC V5A 1S6 OFFICE: SSC K-10549 PHONE : 778-782-8037 E-MAIL: tloughin "at" sfu "dot" ca |
||||
|
What the heck happened in the middle??? |
I actively support principles of equity, diversity, and inclusiveness. I
strongly encourage outstanding statistics undergraduates and Masters of all gender, racial, ethnic, and personal backgrounds to
apply for graduate study in the department and with me. I believe that diversity of
student backgrounds enriches my team, and I welcome the opportunity to work
and engage with all students.
The SECOND EDITION of my book, Analysis of Categorical Data with R, is now
available! The book’s website is . Order a copy directly from the publisher, , or from fine merchants around the world.
My Duties include:
RESEARCH
·
I am interested in developing
Statistical Methodologies in a wide range of areas. Recent work includes:
·
Statistical learning, including regression trees and tree-based ensembles
. Regression trees, as they are currently used---both alone and in
ensembles---are very limited tools. With my graduate students we are developing tree
and ensemble algorithms that adapt better to unexpected data structures and are
more easily tuned than usual methods.
·
Modeling categorical data, particularly the analysis of multiple
response categorical data. This problem arises often in surveys where respondents are told to
“mark all that apply”. Surprisingly, until recently, there were no good,
established methods for modeling data of this type. Chris Bilder
and I have been working on a class of models that are flexible yet easy to
interpret. See .
·
Multi-Label Classification. Multi-label classification (MLC) deals
with the question of how to best classify objects on which a vector of
binary classes is observed. Independent classification of each label ignores
potentially valuable information on the joint distribution of the binary labels.
This problem lies at the intersection of my work on statistical learning and
modeling "choose all that apply" questions.
·
Analysis of unreplicated experiments. When factorial experiments are
performed without replication, the analysis of factors that affect the mean is
tricky. Even trickier is the analysis of factors that affect the variance. In
collaboration with students, I have developed several methods for improving
inferences from unreplicated factorials.
·
Design and analysis of long-term experiments. When field trials and certain other
experiments are run over a long time, they are subject to random effects due to
years. Most experimental designs and analysis methods address this problem very
poorly. With colleagues, we have developed improved designs and are applying them
to actual long-term experiemnts in fisheries research,
·
Collaborative Work. with researchers in various areas. Recent areas of collaboration
include Pediatric Medicine; Cardiovascular Physiology; Wildlife Ecology and
Biology; Interactive Arts and Technology; Laboratory Testing; and others.
·
Quantitative Comedy. Banquet talks and serving as MC at events. I like to mix fun and business.
(References available on request.)
·
A
detailed CV is available here (pdf format) .
STATISTICAL CONSULTING
·
I am available
for statistical consulting and collaboration for ¶¡ÏãÔ°AV researchers as well as external
people and organizations. My internal consulting is mainly limited to collaborations
where publication is expected and a statistical collaborator would be helpful.
Ian Bercovitz at ¶¡ÏãÔ°AV Statistical Consulting Service is excellent and can handle other problems.
Read here
to find out more about the ¶¡ÏãÔ°AV Statistical Consulting Service.
·
The best time to reach out for statistical consulting is in the planning
phase of the study.
Statistical consultation early can help to avert
a disaster much more easily than we can help to recover from one.
·
Don’t make these mistakes!
TEACHING
See my department profile for my current and upcoming course offerings.
My teaching history includes a wide range of courses:
·
Graduate
Level Statistics: Modern Methods in Applied Statistics (Statistical Learning),
Bootstrapping, Linear Models and Messy Data, Multiple Testing and Multiple
Comparisons, Lifetime Data Analysis
·
Upper-Level
Applications Courses: Statistical Learning, Categorical Data Analysis, Design of Experiments, Applied
Linear Models
·
Service
Courses: Intro to Stat, Chance and Data Analysis, ANOVA, Regression, Intermediate
Probability and Statistics, Biostatistics
DOWNLOADS
·
The
SAS/IML program to perform the Loughin and Noble (1997) permutation test for unreplicated factorials is
Details regarding my education, background, experience, and publications
can be found on my
Expanded CV
(pdf)
|
Department of Statistics Faculty ¶¡ÏãÔ°AV Department of Statistics and Actuarial Science Find out what the is... Find out … |
Last Update: 2024 July 04