1 min read

Science and statistical inference

Q: There have been a lot of papers recently about the spike in p-values just below 0.05, haven’t there?

A: Yes. A bit depressing. But there’s a new analysis that says it’s ok.

Q: Really? That’s great!

A: Yes, Daniel Lakens shows that you can explain the recent increase in just-significant p-values, and that “data does not provide any indication of an increase in questionable research practices”

Q: And is his modelling correct?

A: It certainly looks plausible.

Q: So what does explain the 0.04 spike, if it’s not ‘questionable research practices’

A: Let’s see…Ah. A decrease in power and increase in publication bias.






** crickets **






Q: Sorry, I got nothing.

A: Well, better go, grant proposals aren’t going to write themselves.