STATISTICAL SOCIETY OF AUSTRALIA QUEENSLAND BRANCH ANNUAL GENERAL MEETING 2018
4:30PM, TUESDAY 20 MARCH, 2018
VENUE: The Gibson Room 1064, Z Block, Gardens Point Campus, QUT
Followed by an ordinary branch meeting at 5:15PM at the same venue. Non-members welcome, as always. Members and guests are welcome to join the speaker afterwards at a nearby restaurant.
Speaker: Professor Adrian Barnett, Queensland University of Technology
Title: The crisis in health and medical statistics
Abstract: Health and medical research is vital for improving people’s lives. At its best it involves skilled researchers testing relevant questions using rigorous methods that translate into practice. Unfortunately much research fails to improve health because of long-standing issues such as inadequate training, poor regulation and perverse incentives. Recent attempts to reproduce peer reviewed papers in health and medicine had failure rates between 75% and 90%, and an estimated 85% of research is currently lost due to poor research practices.
Poor statistical practices are central to this “reproducibility crisis”. Many health and medical researchers – even those with years of experience – have little understanding of the correct definition of a p-value, confidence interval or confounder. Too many researchers follow the “publish or perish” maxim and salami slice their research and promote “statistically significant” findings that are not reproducible. The bias to publish “statistically significant” results means crucial evidence that fails to show a difference is often never published. Other crucial evidence may be suppressed because of conflicts of interest.
The problem with statistics is not the methods themselves, but in their application. Hence statisticians have a key role in tackling the current crisis by promoting best practice and training their colleagues. We can also study the problems and design evidence-based interventions to reduce them as part of the new field of meta-research or “research on research”. Although the current situation is grim there is enormous potential to increase the value of health and medical research by promoting better statistical practice.
Biography: Professor Barnett has a Bachelor of Science in Statistics from University College London and a PhD from the University of Queensland; his thesis developed new methods to detect non-linearity in time series. He has worked for over 21 years as a statistician, working for drug companies, research councils and universities. He is an NHMRC Senior Research Fellow with a project title of: “Meta-research: Using research to increase the value of health and medical research”. He is the current vice-president of the Statistical Society of Australia.
|Time:||4:30 pm - 6:30 pm|