Moving beyond confirmation – a new article on exploration

Hidden exploration of data fishes through p-hacking and HARKing for apparent evidence of claims, contributing to the replication crisis. Transparent exploration, however, rightly enjoys all of the freedoms of looking for patterns in data to improve hypotheses and models or make new claims. It invites confirmatory studies to test the new claims using new data.

In this second part of "Means to valuable exploration", Michael Höfler, Brennan McDonald, Philipp Kanske und Robert Miller give guidance on how to approach this. On the one hand, one should find all patterns that advance a field of research, but on the other hand, one should reduce the results to the most relevant ones. We give an overview of statistical smoothing and filtering techniques that serve this goal. We also make suggestions to editors of scientific journals and other decision-makers to establish more transparent exploration.

You can read the article here.