Menu
Log in


'The Eye of the Beholder' : Regression Coefficients & Mechanical Objectivity in Public Health Research

  • 21 Jul 2020
  • 6:00 PM - 7:00 PM (UTC+10:00)
  • Online, via Zoom

Registration

  • This is the registration type for all members and non-members.

Registration is closed

SSA Vic is pleased to host a public seminar by Taya Collyer from Monash University.

Abstract:

In this talk, I position current debate regarding 'mechanistic' statistical practice in the context of 18th and 19th century debates regarding scientific image-making and 'Mechanical Objectivity'. I apply concepts from the Sociology of Scientific Knowledge (SSK) and Science and Technology Studies (STS) to 21st century public health research via an empirical investigation of statistical practice within the field of health equity research. Qualitative interview data from 45 academics representing 15 disciplines in 8 countries suggest that, despite their mechanistic origins, the interpretation of regression coefficients (and associated p-values) in applied research contexts depends heavily on the human analyst; their expectations, their suspicions, their personal ‘lens of commonsense’. Disciplinary training forms an important part of this lens, epidemiologists and economists had a more technical focus on interpreting coefficients, while social scientists were more likely to be reflexive, and aware of their own role in the construction of statistical knowledge. However, biostatisticians and other interviewees with advanced statistical training also stressed the important role played by the analyst in shaping statistical results.

In evaluating work produced by others, negative modalities functioned to limit the extent to which regression coefficients are understood to correspond with reality. Duhem's Paradox - a researcher's inability to determine whether unexpected results arise from methodological error or represent discovery- was clearly apparent in discussions of statistical inference, and this challenge cut across disciplinary boundaries, uniting researchers from disparate backgrounds. This may explain why public health has (and many other fields have) been unable to arrive at a satisfactory replacement for the much-maligned ‘Statistical Significance’ concept, which, for all its faults, has currency in a wide range of disciplinary domains. Together, results confirm my suspicion that a sociology of biostatistical practice would usefully contribute to long-standing debates which, presently, appear intractable.


Bio:

Taya Collyer is an early-career biostatistician and sociologist of statistical practice. She completed her Master of Biostatistics at Monash University in 2016 and is in the process of submitting her PhD thesis (from which this presentation is drawn) at the University of Edinburgh, School of Social & Political Science. When not masquerading as a social-scientist, her research focus is the creative generation and analysis of large-scale clinical datasets to answer questions about structural drivers of health, especially in older people. Taya has significant data management expertise, developed during the progress of the ASPREE clinical trial (n=19,114), the largest primary-prevention study of aspirin ever undertaken in healthy older people. Taya is currently appointed as a Research Fellow (biostatistics) in the Academic Unit at Peninsula Health, where she provides methodological support for a range of qualitative and quantitative projects, including the linked data platform and National Centre for Health Aging.

Media:

The presentation is now available in video, at: https://vimeo.com/ssavic/2020721-collyer, and in audio, at: https://bit.ly/3jA0QSg. Slides from the presentation are also available at: https://drive.google.com/file/d/1Ol7SXLHba2rAW20PVrdhcigEFeWf3FH9/view?usp=sharing.

Zoom link:

Link and password to the Zoom session will be provided, by email, upon registration.

Contact:

For further details, do not hesitate to contact the organisers: Emi Tanaka (emi.tanaka@monash.edu) and Elizabeth Korevaar (elizabeth.korevaar1@monash.edu).

Powered by Wild Apricot Membership Software