In this episode, Dan and James welcome back Daniel Lakens (Eindhoven University of Technology) to discuss his new paper on justifying your alpha level.
Highlights:
Why did Daniel write this paper?
Turning away from mindless statistics
Incremental vs. seismic change in statistical practice
The limitations to justifying your alpha
The benefits of registered reports
Daniel’s coursera course
What’s better? Two pre-registered studies at .05 or one unregistered study at .005?
Testing at the start of semester vs. the end of semester
Thinking of controlling for Type 1 errors as driving speed limits
Error rates mean different things between fields
What if we applied the “5 Sigma” threshold used in physics to the biobehavioral sciences?
What about abandoning statistical significance
How did Daniel co-ordinate a paper with 88 co-authors?
Using time zones to your benefit when collaborating
How can junior researchers contribute to these types of discussions?
Science by discussion, not manifesto
The dangers of blanket recommendations
How do you actually justify your alpha from scratch?
Links
Daniel on Twitter - https://www.twitter.com/lakens
Daniel’s courser course - https://www.coursera.org/learn/statistical-inferences
Justify your alpha paper - https://psyarxiv.com/9s3y6
Abandon statistical significance - https://arxiv.org/abs/1709.07588
Using the costs of error rates to set your alpha - https://doi.org/10.1111/j.1461-0248.2004.00625.x Special Guest: Daniel Lakens.