March 8, 2016

A link for the day

By

My first Listener column, a couple of years ago was on the US gender pay gap and what it does and doesn’t mean. It’s now open-access.

avatar

Thomas Lumley (@tslumley) is Professor of Biostatistics at the University of Auckland. His research interests include semiparametric models, survey sampling, statistical computing, foundations of statistics, and whatever methodological problems his medical collaborators come up with. He also blogs at Biased and Inefficient See all posts by Thomas Lumley »

Comments

  • avatar
    Megan Pledger

    I followed the links reading until I got to “Out of Control”. From what you said it kinda occurred to me that the placebo effect is partly regression to the mean. Patients have to meet some sort of illness criteria to get into an RCT but for some people that’s not their typical state but a more ill state at that point in time. A little later they go back to their more typical, healthier state. And it’s counted as the placebo effect.

    The people who would balance them out i.e. average health at baseline but sicker at a later stage (before reverting to their typical state) never get into the trial because they weren’t sick enough when measured at baseline.

    8 years ago

    • avatar
      Thomas Lumley

      Yes, regression to the mean is definitely one of the components of the ‘placebo effect (broad sense)’. It’s not the only component, though. Here’s an example I use in class:

      Imagine testing a cough-suppressant treatment:

      • Medication reduces coughing
      • Placebo effect reduces coughing
      • Regression to the mean: coughing gets better spontaneously
      • Sedative effect makes coughing less troubling
      • Observer bias: coughing is under-recorded because doctor expects it to work
      • Social pressure: patient under-reports coughing to make doctor happy
      • Recall bias: patient expects the treatment to work and so remembers less coughing

      For pain, there’s evidence for a real placebo effect mediated by endorphins because opioid antagonists block (part of) the effect.

      There are also comparisons of the control arms for different trials of gastric ulcer treatments, which have a reasonably objective outcome, where different sorts of placebo seem to have different effects.

      8 years ago