Log in

NSW Branch: Annual Lecture by Prof Ian Marschner

  • 13 Nov 2019
  • 6:00 PM - 7:00 PM (AEDT)
  • Aerial Function Center, UTS

P-values: the good, the bad and the wrong

P-values have long been a mainstay of statistical analysis but have been increasingly subject to criticism and controversy. Contributions to the P-value controversy have ranged from calls for an outright ban on reporting P-values, through to measured consensus statements on the pros and cons, through to more wholehearted defences. In this talk, drawing on some of the recent discussion in the literature (including the American Statistical Society statement) I will consider some of the valid uses of P-values (the “good”) along with some of their invalid and undesirable uses (the “bad”). In addition, I will attempt to debunk some of the invalid arguments that have been presented against P-values (the “wrong”). Overall, I will present a case that P-values are never the most important aspect of an analysis, but they can add some useful supplementary information to other more informative tools, such as point estimates, standard errors and confidence intervals. Whatever one’s position on P-values, it is important to use sound logic and avoid invalid arguments in reaching that position.


Ian Marschner is Professor of Biostatistics at The University of Sydney, in the NHMRC Clinical Trials Centre. He has 30 years of experience as a biostatistician working on health and medical research, particularly involving randomised clinical trials. His research involves new methodological developments in various areas of statistical design, analysis and computation, and has been supported by NHMRC, ARC and industry funding schemes. Professor Marschner has held senior appointments in both academia and industry, including Head of the Department of Statistics at Macquarie University, Director of the Asia Biometrics Centre with the pharmaceutical company Pfizer, and Associate Professor of Biostatistics at Harvard University.

Powered by Wild Apricot Membership Software