Deeper down the rabbit hole of data, analysis, and inference errors and suggestions for digging back out
Speakers: D. Allison
If research is important enough to publish, then David Allison believes we should care enough to get it right. We need to hold our science to a higher level. Just because the measurement method we have to hand is the best available does not make it adequate.
In this presentation, he explains regular errors of research design, errors of analysis and errors in reporting – he gives examples where research has had to be retracted. Statistical quality control is not great in many research environments, he notes, and he lays out why we need to think more mathematically. If this is achieved, then headline distortions in the media will be reduced.
We also need to develop procedures and a culture that better supports expeditious and civil correction of detected errors.