The RSS is a professional body for all statisticians and data analysts wherever they may live.
My RSS
My RSS
Membership
Join now
Member benefits
Fellow
Professional membership
Corporate membership
e-Student
Meet our members
FAQs
Fees
Professional Development
Data Analyst
Graduate Statistician
Chartered Statistician
Data Science Professional
Advanced Data Science Professional
Professional Statistician's Network
Professional Register
Mentoring scheme
Accreditation scheme
Sections and local groups
Sections
Local groups
Special interest groups
Volunteering
Future Leaders
Maths 4 Girls
Statisticians for Society
Volunteer with the African Institute for Mathematical Sciences (AIMS)
Volunteer profiles
Find a volunteering opportunity
Volunteer as a charity trustee
Other pro bono opportunities
Guidelines for RSS volunteers
Celebrating statisticians
Medals and prizes jointly awarded by the RSS
RSS honours
Excellence awards
External honours panel
Statistics Are Vital
News and Publications
News
Section and group news
Member callouts
Press and media
Publications
Journals
Significance
Real World Data Science
Statistics law guides
Our research
Change in publisher in 2023
Training and events
Training
Public courses
Online courses
In-company courses
Lean & 6 sigma
CPD
Stats resources
Conference 2025
Accessibility information
Conference costs
Conference streams
Grants and bursaries
Invited topic sessions
Partnership opportunities
Registration
Submit a talk or poster
Terms and conditions
Venue, accommodation and travel
Conference 2024
Programme and speakers
Past conferences
Events
Events calendar
Video and audio recordings
President's Address
Excellence Awards
RSS Honours
Discussion paper meetings
Named lectures
Webinars
Policy and campaigns
Policy
Our asks of government
Our work on public statistics
Our Data Manifesto
Statistics Under Pressure
Upcoming policy consultations
Climate change resources
Our work on Covid-19
Our external affairs strategy
Policy groups
AI Task Force
Campaigns Advisory Group
Academic Affairs Advisory Group
Covid-19 Task Force
Climate Change and Net Zero Task Force
Data Science Task Force
Education Policy Advisory Group
National Statistics Advisory Group
Statistics Under Pressure Steering Group
Working Group on Diagnostic Tests
Search policy documents
Covid-19
Data science (strategy, ethics and skills)
Education
Government statistics
Higher education, research and innovation
Stats for the public good
Jobs and careers
Jobs board
Advertise with us
Career development
Opportunities at the RSS
Resources
Resources for journalists
Online courses for journalists
How to write numbers
Teaching resources
Hands-on statistics
Teaching and outreach resources
e-Teacher membership
Statistical explainers
Advice when looking at stats
Climate change explainers
Covid-19 FAQs
Find an expert
Statistical ambassadors
Professional register
Consultants Directory
Audio and video
Sign in
Join
Introducing the RSS
A short, two minute video introduction
Watch our video
Who we are and what we do
Our history and where to find us
View our 'About' section
Consultants Directory
Find a professionally qualified statistical consultant
Find a consultant
President and staff
Meet our personnel
Meet our president and staff
About RSS
Home
Events
Events 2025
Local Groups
Reducing Multicollinearity in GLMs with categorical covariates
Reducing Multicollinearity in GLMs with categorical covariates
Date:
Wednesday 16 April 2025, 1.00PM
Location:
Queens University Belfast, Maths & Physics Teaching Centre Room G005 -between the Admin Building and the Library
Local Group Meeting
Share this event
Event overview
Full event information
Speakers
Organiser
On Wednesday 16th of April, at 13:00, there
will be a talk
in
the Maths and Physics Teaching Centre (Ground floor Room 005),
QUB,
by Gilbert MacKenzie, formerly of the University of Limerick (ROI) and ENSAI (France). This will be a hybrid event with the speaker in-person and on-line using MS teams
(link).
Abstract:
When dealing with GLMs with categorical covariates we show that varying the reference subclasses leads to different variance-covariance matrices and develop a relation between a measure of precision and a measure of multi-collinearity, by analysis and by simulation. The net result is that we are able to demonstrate, inter alia, how to reduce multi-collinearity by a judicious choice of reference subclasses in GLMs with categorical covariates. We develop estimators for the discrete choice minima of our measures and evaluate their performance in a wide class of GLM structures likely to arise in practice. Our approach resolves some queries which have arisen in the literature.
All welcome!
See previous talks
here.
Read previous write-ups
here
.
Follow RSSNI on
Linkedin
Gilbert MacKenzie (Chair) &
Hannah Mitchell (Meetings Host)
RSSNI
Professor Gilbert MacKenzie for Northern Ireland Local Group
Contact
Prof. Gilbert MacKenzie